TW453080B - Method and apparatus for selectively discarding packet for blocked output queues in the network switch - Google Patents

Method and apparatus for selectively discarding packet for blocked output queues in the network switch Download PDF

Info

Publication number
TW453080B
TW453080B TW087101988A TW87101988A TW453080B TW 453080 B TW453080 B TW 453080B TW 087101988 A TW087101988 A TW 087101988A TW 87101988 A TW87101988 A TW 87101988A TW 453080 B TW453080 B TW 453080B
Authority
TW
Taiwan
Prior art keywords
frame
port
copy
queue
output
Prior art date
Application number
TW087101988A
Other languages
Chinese (zh)
Inventor
Bahadir Erimli
Thomas J Runaldue
Chandan Egbert
Denise Kerstein
Original Assignee
Advanced Micro Devices Inc
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Advanced Micro Devices Inc filed Critical Advanced Micro Devices Inc
Application granted granted Critical
Publication of TW453080B publication Critical patent/TW453080B/en

Links

Classifications

    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L49/00Packet switching elements
    • H04L49/10Packet switching elements characterised by the switching fabric construction
    • H04L49/103Packet switching elements characterised by the switching fabric construction using a shared central buffer; using a shared memory
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L12/00Data switching networks
    • H04L12/02Details
    • H04L12/16Arrangements for providing special services to substations
    • H04L12/18Arrangements for providing special services to substations for broadcast or conference, e.g. multicast
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L41/00Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
    • H04L41/02Standardisation; Integration
    • H04L41/0213Standardised network management protocols, e.g. simple network management protocol [SNMP]
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L49/00Packet switching elements
    • H04L49/20Support for services
    • H04L49/201Multicast operation; Broadcast operation
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L49/00Packet switching elements
    • H04L49/25Routing or path finding in a switch fabric
    • H04L49/253Routing or path finding in a switch fabric using establishment or release of connections between ports
    • H04L49/255Control mechanisms for ATM switching fabrics
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L49/00Packet switching elements
    • H04L49/30Peripheral units, e.g. input or output ports
    • H04L49/3027Output queuing
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L49/00Packet switching elements
    • H04L49/35Switches specially adapted for specific applications
    • H04L49/351Switches specially adapted for specific applications for local area network [LAN], e.g. Ethernet switches
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04QSELECTING
    • H04Q3/00Selecting arrangements
    • H04Q3/42Circuit arrangements for indirect selecting controlled by common circuits, e.g. register controller, marker
    • H04Q3/54Circuit arrangements for indirect selecting controlled by common circuits, e.g. register controller, marker in which the logic circuitry controlling the exchange is centralised
    • H04Q3/545Circuit arrangements for indirect selecting controlled by common circuits, e.g. register controller, marker in which the logic circuitry controlling the exchange is centralised using a stored programme
    • H04Q3/54575Software application
    • H04Q3/54591Supervision, e.g. fault localisation, traffic measurements, avoiding errors, failure recovery, monitoring, statistical analysis
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L12/00Data switching networks
    • H04L12/02Details
    • H04L12/16Arrangements for providing special services to substations
    • H04L12/18Arrangements for providing special services to substations for broadcast or conference, e.g. multicast
    • H04L12/1863Arrangements for providing special services to substations for broadcast or conference, e.g. multicast comprising mechanisms for improved reliability, e.g. status reports
    • H04L12/1868Measures taken after transmission, e.g. acknowledgments
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L49/00Packet switching elements
    • H04L49/20Support for services
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L49/00Packet switching elements
    • H04L49/30Peripheral units, e.g. input or output ports
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L49/00Packet switching elements
    • H04L49/30Peripheral units, e.g. input or output ports
    • H04L49/3009Header conversion, routing tables or routing tags
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L49/00Packet switching elements
    • H04L49/30Peripheral units, e.g. input or output ports
    • H04L49/3054Auto-negotiation, e.g. access control between switch gigabit interface connector [GBIC] and link
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L49/00Packet switching elements
    • H04L49/35Switches specially adapted for specific applications
    • H04L49/354Switches specially adapted for specific applications for supporting virtual local area networks [VLAN]
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L49/00Packet switching elements
    • H04L49/45Arrangements for providing or supporting expansion
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04QSELECTING
    • H04Q2213/00Indexing scheme relating to selecting arrangements in general and for multiplex systems
    • H04Q2213/1316Service observation, testing
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04QSELECTING
    • H04Q2213/00Indexing scheme relating to selecting arrangements in general and for multiplex systems
    • H04Q2213/13162Fault indication and localisation

Abstract

An apparatus and method of controlling the transmission of copies of a frame of data from a multiport device, such as a mulitport network switch, determines which ones of the ports are designated to transmit a copy of a multicopy frame received at the switch. The switch then determines which of these designated ports are not available to transmit a copy of the frame. The copies are transmitted from those designated ports that are available to transmit a copy of the multicopy frame, while the copies are discarded for those designated ports that are not available to transmit a copy of that multicopy frame.

Description

4 ο 8〇 Β7 經濟部中央標準局員工消費合作杜印製 五、 發明説明 1 | 發明之 領— 域 1 | 本 明 是 Η 關 於 通 訊 的 領 域 > 特 別 是 有 關 於 由 裝 置 的 1 I 多 m 埠 (P 0 Γ t ) 傳 輸 多 個 資 料 副 本 之 方 法 及 裝 置 〇 二 請 1 1 先. 1 發—明—背—景.. 一 讀 背 1 1 在 許 多 網 路 系 統 中 > 例 如 封 包 交 換 網 路 (即乙太網路) 面 I 之 1 | 1 經 常 需 要 由 個 單 端 網 站 傳 送 相 同 的 資 料 到 複 數 的 終 端 意 事 項 1 I 網 站 (S t R t 1 on )〇 例如 傅铳的e m 1Π i系統, -俩使用者 存 1 1 Η 本 1 裝 需 要 送 相 同 的 e - m a ί 1 訊 息 給 連 到 此 e〜 m a i 1 Ά 統 中 的 另 外 四 頁 1 個 使 用 者 〇 1 I 在 一 封 包 交 換 網 路 中 一 個 交 換 裝 置 當 它 從 交 換 裝 置 \ 1 I 埠 接 收 一 個 來 白 另 一 终 端 網 站 的 訊 框 (f r a me S of d a t a )時 1 i 訂 * 要 做 轉 送 的 決 策 〇 如 果 訊 框 被 送 到 多 個 終 端 網 站 9 則 此 1 交 換 裝 置 必 須 送 此 訊 框 的 副 本 到 複 數 個 正 確 的 埠 〇 1 1 在 幾 乎 大 多 數 的 系 統 褢 > 存 在 許 多 系 統 的 來 源 端 過 分 1 1 負 荷 的 例 子 〇 例 如 » 在 _a 個 具 有 多 個 埠 { P 0 Γ U的 網 路 4 在 1 線 某 些 特 定 的 — 或 數 個 埠 t 會 有 向 外 交 通 擁 塞 的 情 形 > 使 得 1 I 瑄 些 埠 至 少 暫 時 無 法 用 來 傳 送 資 料 〇 但 其 他 埠 仍 可 用 來 傳 1 1 | 送 資 科 〇 因 此 * 有 些 被 指 定 用 來 傳 送 同 樣 複 製 資 料 的 埠 被 1 1 塞 住 了 * 其 他 被 指 定 用 來 傳 送 複 製 資 料 埠 則 沒 有 塞 住 0 1 1 傳 統 的 多 埠 網 路 交 換 裝 置 f 當 所 有 被 指 定 用 來 傳 送 複 1 製 資 料 埠 都 被 塞 住 時 9 則 該 要 被 傳 送 的 資 料 會 被 丢 棄 > 同 i 1 時 該 被 丟 的 資 料 會 通 知 管 理 % 統 〇 只 有 些 指 定 的 埠 被 塞 1 | 住 而 其 他 沒 有 的 情 況 下 ί 該 丟 棄 的 程 序 仍 會 執 行 〇 換 句 話 i I 說 在 某 lih 揷 情 況 下 • 所 有 的 資 料 會 被 丟 棄 » 而 沒 有 複 製 1 1 本紙張尺度適用中國國家標準(CNS ) Λ4規格(210 X 297公釐) 1 9 1472 4 53 ο4 ο 8〇Β7 Staff Consumer Cooperation of the Central Bureau of Standards, Ministry of Economic Affairs, Du Du, V. Invention Description 1 | Field of Invention-Domain 1 | This is clearly Η About the field of communication > Port (P 0 Γ t) Method and device for transmitting multiple copies of data. 2 Please 1 1 first. 1 send-Ming-back-scene. 1 read back 1 1 in many network systems > such as packet switching network Road (ie, Ethernet) 1 of 1 | 1 It is often necessary for a single-ended website to send the same data to a plurality of terminals. 1 I website (S t R t 1 on). For example, Fu Yi's em 1Π i System, two users save 1 1 Η this 1 device needs to send the same e-ma ί 1 message to the other four in the e ~ mai 1 Ά system 1 user 〇1 I a switching device in a packet switching network when it receives a frame (frame S of data) from the switching device \ 1 I port to the other terminal website 1 i order * Make forwarding decisions. If the frame is sent to multiple terminal sites, then the exchange device must send a copy of the frame to the correct ports. 1 In most systems, there are many systems. Example of excessive 1 1 load on the source side. For example »In _a networks with multiple ports {P 0 Γ U 4 On line 1 some specific — or several ports t will have traffic congestion outwards> ; Make 1 I some ports ca n’t be used to transmit data at least temporarily; but other ports can still be used to transmit 1 1 | The data port is blocked by 1 1 * Other ports designated for transmitting duplicate data are not blocked 0 1 1 Traditional multi-port network switching devices f When all the ports designated for transmitting duplicate data are blocked At the time of live 9, the data to be transmitted will be discarded > The same as i 1, the data to be lost will be notified to the management% System 0 Only some designated ports are blocked 1 | Live and others are not The procedure will still be executed. In other words, i I said that in a certain case of lih 揷 • All materials will be discarded »without copying 1 1 This paper size applies the Chinese National Standard (CNS) Λ4 specification (210 X 297 mm) 1 9 1472 4 53 ο

8C Α7 Β7 經濟部中央標準局員工消費合作社印製 五、 發明説明 :2 ) 1 | 資 料 送 出 即 使 只 有 一 個 埠 擁 塞 而 十 個 埠 沒 有 〇 這 種 程 序 1 1 會 造 成 在 壅 塞 的 時 候 * 相 對 大 量 的 資 料 被 丟 棄 1 因 為 只 要 1 I 一 個 指 定 的 埠 被 塞 住 了 9 將 阻 擋 住 提 供 資 料 到 剩 下 的 埠 的 請 先 1 1 1 資 料 叫 I 讀 背 1 之摘要_ I 之 1 裝 置 注 I 因 此 爾 要 安 排 及 使 用 棰 方 法 t 用 來 控 制 由 個 意 事 項 再 1 1 之 多 個 埠 傳 送 複 製 的 資 料 0 使 該 裝 置 較 少 丟 棄 資 料 而 傳 送 1 1 填 1 更 多 的 資 料 0 寫 本 裝 頁 1 此 種 需 求 和 其 他 需 求 本 發 明 用 — 個 實 施 例 來 滿 足 〇 1 I 本 發 明 提 出 - 個 方 法 用 來 控 制 由 多 個 埠 送 出 的 複 製 訊 框 ( 1 | C 〇 p y 0 f f Γ a m e ) > 該 方 法 並 指 定 一 些 埠 f 訊 框 將 由 此 些 埠 1 訂 送 出 4 然 後 此 方 法 會 決 定 哪 些 埠 不 用 來 傳 輸 訊 框 〇 將 要 傳 1 送 的 訊 框 提 供 到 這 可 用 來 傳 送 複 製 訊 框 中 的 一 涸 指 定 埠 1 1 0 然 而 這 些 可 用 來 傳 送 複 製 訊 框 的 指 定 埠 並 不 用 來 傳 送 被 1 1 丢 棄 的 複 製 訊 框 0 1 旅 這 種 方 法 的 優 點 是 9 複 製 訊 框 只 由 指 定 的 埠 傳 m > 且 1 I 這 些 埠 都 能 傳 送 資 料 〇 唯 --* 會 被 丟 棄 的 複 製 訊 框 * 是 那 些 1 1 I 將 要 由 指 定 的 埠 傳 送 的 複 製 訊 框 卻 因 某 種 原 因 而 不 能 被 1 1 傳 送 〇 因 此 比 單 純 地 只 是 丟 棄 整 個 訊 框 * 本 發 明 提 供 了 1 1 由 埠 到 终 點 平 台 更 多 的 傳 輸 容 量 0 因 此 , 一 m 塌 的 擁 塞 並 i 1 不 會 m 擋 提 供 到 如 此 多 的 指 定 埠 的 資 料 C 因 此 訊 框 資 料 將 1 1 儘 可 能 由 那 些 想 接 收 訊 框 的 接 收 者 接 收 〇 1 I ·» r. 月U 述 的 需 要 1 也 由 本 發 明 之 另 一 實 施 例 提 出 一 個 具 多 1 | 個 埠 能 傳 輸 m ~ 訊 框 之 多 假 副 本 的 網 路 父 換 裝 置 t 來 滿 1 1 本紙張尺度適用中國國家標準(CNS ) A4規格(210X297公釐) 91472 τ)8C Α7 Β7 Printed by the Consumer Cooperatives of the Central Standards Bureau of the Ministry of Economic Affairs 5. Description of the invention: 2) 1 | Data submission Even if only one port is congested and ten ports are not. This procedure 1 1 will cause a relatively large amount of congestion * Data is discarded 1 because as long as 1 I a designated port is blocked 9 will block the supply of data to the remaining ports please first 1 1 1 The data is called I read the summary of 1_ I 1 device Note I Therefore It is necessary to arrange and use the method t to control the transmission of duplicated data from multiple ports of individual items 1 1 0 to make the device less discard the data and send 1 1 fill 1 more data 0 write this page 1 Needs and other needs The present invention uses an embodiment to satisfy 〇1 I. The present invention proposes a method for controlling a copy frame (1 | C 〇 0 0 ff Γ ame) sent from multiple ports > This method also specifies some ports f The frame will be ordered to send out 4 from these ports 1. Then this method will determine which ports are not used to transmit the frame. 0 The frame to be transmitted is provided to the designated port 1 1 0 which can be used to transmit the duplicate frame. However these The designated port that can be used to send the duplicate frame is not used to transmit the duplicate frame that was discarded by 1 1 0 1 The advantage of this method is that 9 duplicate frames are transmitted only by the designated port m > and 1 I these ports are Can transmit data 〇 Only-* duplicate frames that will be discarded * are those 1 1 I duplicate frames that will be transmitted by the designated port but cannot be transmitted by 1 1 for some reason. Therefore, it is more than just discarding the entire Information Frame * The present invention provides 1 1 more transmission capacity from the port to the destination platform. 0 Therefore, one m of congestion and i 1 will not provide data to so many designated ports. Therefore, the frame data will be 1. 1 Received by those who want to receive the frame as much as possible. 〇1 I · »r. Needs described in month U. Another embodiment of the present invention also proposes that one port can transmit m ~ frame as many as possible. False copy of the network parent replacement device t to full 1 1 This paper size applies the Chinese National Standard (CNS) A4 specification (210X297 mm) 91472 τ)

453〇 S A7 B7 經濟部中央標準局員工消費合作社印製 V♦ 五、 發明説明 ) 足 此 擗 要 〇 該 網 路 交 換 裝 置 具 有 多 個 埠 » 由 這 些 埠 傳 送 複 製 訊 框 〇 一 個 埠 向 量 產 生 器 (P or t v e c t 0 Γ ε e η e r at 〇 r )用 來 產 生 埠 向 量 該 埠 向 最 抱 定 由 那 一 個 埠 傅 送 複 製 訊 框 0 一- 個 傳 輸 的 規 劃 決 定 瑄 些 被 埠 向 量 (P 0 Γ t ν e c t or )指定傳 輪 單 一 個 複 製 訊 框 埠 之 利 用 率 〇 該 傳 輸 規 制 使 得 複 製 訊 框 被 指 定 傳 輪 複 製 訊 框 的 埠 中 可 利 用 的 埠 所 傳 輸 4 丟 掉 複 製 訊 框 的 動 作 1 則 由 被 指 定 傳 輸 複 製 訊 框 的 埠 中 不 能 利 用 的 埠 來 做 〇 本 發 明 上 述 實 施 例 之 優 點 為 f 該 父 換 裝 置 儘 可 能 的 在 網 路 中 傳 輸 最 多 的 複 製 訊 框 〇 如 果 只 是 少 數 幾 個 埠 沒 被 利 用 於 傳 輸 訊 框 » 並 不 會 白 動 地 丟 棄 複 製 訊 框 0 如 此 增 加 網 路 中 有 用 資 料 量 的 傳 輸 避 免 第 —- 次 傳 輸 訊 框 到 網 路 父 換 裝 置 » 卻 轉 送 了 —^- 個 全 然 的 網 路 資 源 垃 圾 〇 上 述 的 特 點 及 其 他 別 的 特 點 > 覼 點 及 本 發 明 的 優 點 t 經 下 面 連 结 附 圖 的 詳 细 說 明 , 將 會 更 加 清 楚 〇 _ 式 簡 單 說 明 第 1圖是根據本發明實_例架構之封包交換系铳方塊 _ 〇 第 2圖是根據本發明竄施例之多埠交換裝置方塊圖, 並 被 使 用 在 第 —+ 圖 的 交 換 系 統 中 〇 第 3圖是第二圖根據本發明實施例之多埠交換裝置次 交 換 系 铳 之 圖 示 說 明 0 第 4圖是第三圖根據本發明實施例之次交換系統的- - 涸 輸 出 佇 列 〇 本紙張尺度適用中國國家標準(CNS) Λ4規格(210X 297公釐) 3 91472 經漭部中央標準局員工消费合作社印裝 4 53 〇 , ' Α7 Β7五、發明説明(4 ) 第5圖是根據本發明實腌例之第一種輪出佇列的詳细 结構。 第6圖是根撺本發明實施例之第二種輪出佇列的詳细 结構。 第7圖是根棟木發明實胞例外部記憶體過載區的詳细 结構。 第8圖是本發明中運作之鐽结串列資料结構方塊圖。 第9_是根搪本發明實胞例,示意式地描述訊框鍰衝 器格式q 第10_為依煦第4圖之本發明實施例所建構之交換裝 置次系铳之多副本、回收及空間鍰衝槽區之詳细圖示。 第11圖是根撺本發明實腌例所架構,空閒鍰衝槽结構 的方塊圖。 第12圖是根撺本發明實狍例所配置,多複製佇列之方 塊圖。 第13_是根據本發明實胞例所架構,多複製快取的示 意表示圖。 第14國是根撺本發明一個實胞例所架構,交換裝置糸 統緩衝器管理員排列方塊之方塊圈,及一個FIFO。 示<11L施..例之詳细說明 本發明將以封包交換網路,例如以太網路,中之一交 換裝置為例子來說明。然而很明顯地,本發明亦可應用在 其他如下所述的封包交換系統。 第1圖是本發明的一個範例系铳的方塊圖,在此系統 —I I m —a^^i ^^^^1 ^ϋρ— - —--i Hal· ^^^^1 - ' 、-tf (誚先閱讀背而之注意事項再填艿本頁) 本紙張尺度適用中國國家標準(CNS ) A4規格(210X297公釐) 4 9 1472 53::' A7 B7 五、發明説明(5 ) 中,在操作上將有很大的方便性。本範例系統10是一傾封 包交換系統如K太網路。該封包交換網路包括整合多埠交 換裝貿12,該多埠交換裝賈可致能網路網站間資料封包的 交換。該網路包括不同形態的網路網站,例如,有Μ每秒 1 0百萬位元傳送及接收育料的2 4俩1 0 H \ b P S網路網站1 4 , Μ及每秒1 0 0百萬位元傳送及接收資料的2個1 0 0 M \ b p s網 路網站22。因此,交換裝置12根據以太通訊協定,由網站 1 4或網站2 2接收資料封包,選擇性地轉送資料封包到適當 的目的地。 經墒部中夹標"局負工消費合竹社印繁 (誚先閱讀背而之注意事項再填艿本頁) 根據本掲露之實施例,10H\bps網路網站14,根撺半 雙工的Μ太通訊協定,經介質17由交換裝賈12接收資料或 傳送資料至該交換裝置。以太通訊協定IS0/IEC8802-3( ANSI/IEEE Std.802.3,1993Ed.)定義一個半雙工介質存取 櫬構允許所有的網站14平等地存取網路通道。在介質17上 資料的傳輸並沒有特別加Ml分及具有優先權,而每一個 網站14包括一塊乙太介面卡,該介面卡使用碰撞偵測,載 波感測多重存取(carrier-sense multiple access with collision detection CSMA/CD)來聽(listen)介質上的 資料傳輸。藉著感測介質上接收載波的解除確認( d e a s s e r t i ο n )而偵測到網路資料傳輸的消失。任一個有資 料要傳送的網站14,將在介質上接收載波的解除確認之後 ,等待一個前定時間(predetermined time),再曾試著存 取該通道。該段時間稱為封包間隙時段(interpacket gap interval I P G )。如果複數個網站1 4有資料要在網路上傳 本紙張尺度適用中國國家標準(CNS ) Λ4規格(210X297公釐) 5 9 1472453〇S A7 B7 Printed by the Consumer Cooperatives of the Central Standards Bureau of the Ministry of Economic Affairs v. V. Description of the invention) It is essential that the network switching device has multiple ports »Copy frames are transmitted by these ports. One port vector is generated The device (P or tvect 0 Γ ε e η er at 〇r) is used to generate the port vector. The port is most determined by which port to send a copy frame. 0-a transmission plan determines some of the port vectors (P 0 Γ t ν ect or) Utilization of a single copy frame port in the designated round. This transmission rule makes the copy frame transmitted by the available port in the designated round copy frame. 4 The action of dropping the copy frame 1 It is done by a port that cannot be used among the ports designated for transmitting the replication frame. The advantage of the above embodiment of the present invention is that the parent switching device transmits as many replication frames as possible in the network. If it is only Several ports are not used to transmit frames »and will not discard duplicate frames 0 so that the amount of useful data in the network is transferred to avoid the first-time transmission frame to the network parent switch device» but forwarded — ^-A complete waste of network resources 〇 The above features and other features > Points and advantages of the present invention t will be clearer with the detailed description of the attached drawings below. Fig. 1 is a block diagram of a packet switching system according to an example structure of the present invention. Fig. 2 is a block diagram of a multi-port switching device according to an embodiment of the present invention, and is used in the switching system of Fig. ++. FIG. 3 is a diagram illustrating the second switching system of the multi-port switching device according to the embodiment of the present invention. FIG. 4 is a diagram of the third switching system according to the embodiment of the present invention. This paper size applies the Chinese National Standard (CNS) Λ4 specification (210X 297 mm) 3 91472 Printed by the Consumers 'Cooperative of the Central Standards Bureau of the Ministry of Labor 4 53 〇,' Α7 Β7 V. Description of the invention (4) Figure 5 is based on The detailed structure of the first rotation queue of the actual pickled example of the present invention. Fig. 6 is a detailed structure of a second rotation train according to the embodiment of the present invention. Fig. 7 shows the detailed structure of the external memory overload area of the case of Gendongmu's invention. FIG. 8 is a block diagram of a structure of a serial data structure operating in the present invention. Figure 9_ is a real-life example of the present invention, which schematically describes the frame punch format q Figure 10_ is the multiple copies and recovery of the secondary system of the switching device constructed according to the embodiment of the present invention shown in Figure 4 And a detailed illustration of the space and channel area. Fig. 11 is a block diagram of the structure of a free-form scouring tank constructed according to the actual pickled example of the present invention. Fig. 12 is a block diagram of a multi-copy queue according to the configuration of the embodiment of the present invention. 13th is a schematic representation of a multi-copy cache structured according to the real-life example of the present invention. The fourteenth country is based on the structure of a real cell example of the present invention, the switching device system buffer manager arranges the block circle, and a FIFO. A detailed description of the < 11L application example > The present invention will be described using a packet switching network, such as an Ethernet circuit, as an example. However, it is obvious that the present invention can also be applied to other packet switching systems as described below. FIG. 1 is a block diagram of an exemplary system of the present invention. Here, the system—II m —a ^^ i ^^^^ 1 ^ ϋρ—-—-- i Hal · ^^^^ 1-',- tf (Please read the precautions before filling in this page) The paper size applies to the Chinese National Standard (CNS) A4 (210X297 mm) 4 9 1472 53 :: 'A7 B7 5. In the description of the invention (5) , There will be great convenience in operation. The example system 10 is a packet-switching system such as a Ethernet network. The packet switching network includes an integrated multi-port switching device 12 which enables the exchange of data packets between web sites. The network includes different forms of web sites. For example, there are 24 million megabits per second to send and receive breeding materials. 2 4 H 1 0 H \ b PS web sites 1 4, M and 10 per second. 0 million bits of 2 1 0 M \ bps internet sites 22 for sending and receiving data. Therefore, the switching device 12 receives the data packet from the website 14 or the website 22 according to the Ethernet communication protocol, and selectively forwards the data packet to an appropriate destination. Bid in the Ministry of Economic Affairs " Consumption by the Bureau of Work and Consumption and Printing of the Bamboo Society (诮 Please read the precautions and fill out this page first) According to the example of this disclosure, 10H \ bps Internet website 14, The half-duplex M-tooth communication protocol receives data from the switching device 12 via the medium 17 or transmits data to the switching device. The Ethernet communication protocol ISO / IEC8802-3 (ANSI / IEEE Std.802.3, 1993Ed.) Defines a half-duplex media access architecture that allows all websites 14 to access the network channel equally. The data transmission on the medium 17 does not have special M1 points and has priority, and each website 14 includes an Ethernet interface card that uses collision detection and carrier-sense multiple access with collision detection CSMA / CD) to listen to (data transfer) on the medium. Deactivation of the received carrier on the sensing medium (d e a s s e r t i ο n) detects the disappearance of network data transmission. Any website 14 that has information to transmit will wait for a predetermined time after receiving the de-acknowledgement of the carrier on the medium, and tried to access the channel again. This period of time is called the packet gap interval (IPG). If multiple websites 1 4 have data to upload on the Internet, this paper size applies the Chinese National Standard (CNS) Λ4 specification (210X297 mm) 5 9 1472

53()S Λ7 B7 _ 五、發明説明(6 ) 送,在感測介質上接收載波的解除確認,及等待封包間隙 時段之後,將會嘗試著傳送貴料,且會造成碰撞。因此傳 送資料的網站將監視該介質是杏有因兩網站要同畤傳送而 發生的碰撺r如果偵測到一個碰撞,該兩涸網站都會停止 傳送資料,等待一㈣亂數時間後再重新嘗試傳送。 ΙΟΟΜ/bps的網站22,根據以太標準IEEE802.3X具流最 控制的全雙工工作文件(0.3)所建議,Μ全雙工的棋式操 作會比較好。全雙工的環境提供一介於每一 100 M/bps網 路網站22與交換裝置12間兩路(two way)、點到點的通訊 鐽路,其中交換裝置12及網站22可同時地傳送、接收資料 而不發生碰撞。該100M\bps網站22各別W 100Base-TX, 100Base-T4, 100Base-FX的 100H\bps實體裝置(PHY)2〇8S 接到網路介質17。該交換裝置12包括一介質獨立介面(Mil )24,賴W連接到實體裝置20。 100M\bps的網站22可K用 伺眼器或路由器的形式操作而連接到別的網路。 蛵浐部中央標^^貝工消费合作社印^ (請先閱讀背面之注意事項再填艿本頁) 如第1圖所示,網路10包括一串列的交換裝置送收器 (transceiver)16用執行傳输於交換裝置12及10M\bps網 站14間資料封包的時間分割多工(time division ninlUpiexing)及時間分割解多工。一個磁性變壓器棋組 19維持介質17上的信號波形。交換裝置12包括一送收器介 面18,該送收器介面18用來傅送及接收由交換裝置送收器 16來的資料封包,並在串列N(?Z(series non-return to zero)介面23上使用時間分割多工通訊協定來完成這種工 作。該交換裝置送收器16接收由串列HRZ 23來的封包,解 本紙張尺度適用中园國家標準(CNS ) Λ4現格(210X297公漦) 6 91472 4 53 〇 A7 B7 4£滴部中央標準局貝工消贽合作,社印製 五、發明説明(7 ) 多工該已接收到的封包,並經網路介質1 7輸出該封包到適 當的终站網站14。根據本揭露之實腌例,每一交換裝置送 收器1 6有四個獨立的姣線對(t w i s t e d p a i r )埠,並經N R Z 介面使用4 M的多工,因此整體的接腳比由交換裝置1 2直 接接出減少4倍的接腳。 交換裝置12具有一决策機構(decision making engine), —交換機構(switching engine), —緩衝記憶 體介面,配置/控制/狀態暫存器,管理計數器( management counter) R Μ Λ C (in e d i a access control 媒 體存取控制)通訊協定介面用來支援以太埠網站14、22間 的資料封包路由(routing)。交換裝置12同時也具智慧切 換決策的加強功能,提供Μ管理資訊基礎 (HIB management information base ) 物件型態的統.計網路資 訊給外部的管理賁體,此將敘述於後。交換裝置12也有介 面用來致能外部資料封包的貯存,及交換埋輯(switching Ugic)M降低交換裝置12的晶片尺寸。例如,交換裝置12 具有同步動態RAM(SDRAM)介面34供外郤記憶體36存取,外 部記憶體36用來貯存已接收的訊框、記憶體結構、及MIB 計數器資訊,該外部記憶體36可Μ是具2到4Mb記憶容量的 80、100、1 20HHZW SDRAM〇 交換裝置12同時也包括管理埠30,該管理埠30致能 外部管理資體使該實體藉著管理M A C介面3 2 ( m a g e m e n t MAC inter· fae 32)可控制交換裝置12整體的操作。交換裝 置1 2 ft包话一 P C I介面2 (5 ,用來致能經由P C I主機及橋接 (誚先間讀背而之注意事項再項艿本頁) 本紙張尺度適用中國囤家標率(CNS ) Λ4規格(210XW7公麓) 7 91472 *53 os。 經浐部中央標率局員工消资合作印1;! Λ7 B7 五、發明説明(8 ) 器2 8的外部管理背體存取,另一方面,PCI主機及橋接器 2 8若在有複數個交換裝置1 2的場合,將如掮充確流排般地 使用n 交換裝置12包括一内部決策機構(第二圖),該内部決 策機構甩來選擇性地傳送由住一資料來源來的資料封包到 至少一個目的網站。該内部決策機構亦可以用一個外部規 則檢査器(「u 1 e c h e c k e r )來代替。交換裝置1 2包括一個夕卜 部規則檢査器介面(E R I C ) 4 0 ,使外部規則檢察器可產生一 個訊框轉送決策(frame forwarding dec〖sion)M代替内 部決策機構。因此,框轉送決策可由内部交換機構或由夕卜 部規則檢査器來產生。 交換裝置12也包括一 LED介面44用來輸出每一埠的狀 態及趨動LED外部缠輯46,該LED外部理輯46依次地趨動人 類可讀取的LED元件48。一涸振盪器38提供40MHZ的時脈 輪入給交換裝置12。 第2_是第一_整體多埠交換裝置12的方瑰圖。交換 裝置12包括24個10M\bps媒體存取控制博(media access control port)50,用來在各別的 10M\bps網站(増 1-24) 間,Μ半雙工的方式傳送及接收資料封包。2個100M\bps HAC埠53,用來在各別的l〇〇M\bps網站(埤25、 26)間,Μ 全雙工的方式傳送及接收資料封包。如上所述,管理介面 30也根據MAC麿協定來操作(埠0)。每一 MAC埠50、 53、及 30都有一個接收先入先出(FIFO)媛衝器52及傳送FIF054, 由網站來的資料封包由相對應的MAC埠接收,並儲存在相 I 1^^^* ^^^^1 ^^^^1 —^n tLH -J. (誚先k讀背而之注意事項再填巧本頁) 本紙張尺度適用中國國家標準(CNS ) Λ4規格(210X 297公麓) 8 91472 4 53〇 A7 B7 經^:-部中央標準局負工消費合作社印繁 五、發明説明 ( 9 ) 1 對 應 的 F I F0 52 0 已接收的資料封包由相對應的F IF0 52輸 I 出 到 外 部 J·—t an 憶 體 介 面 34 , 以 便 貯 存 到 外 部 記憶 體 36 0 1 1 已 接 收 封 包 的 表 頭 (h e a d e r ) 同 時 送 到 決策 機 構 « 不 管 1 1 先 是 内 部 規 則 檢 査 器 58 或 外 部 規 則 檢 查 介 面 40 , Μ 決 定 將 由 閱 讀 I 哪 一 傾 Μ A C埠輪出資料。 特別地, 該封包表頭送到内部規 背 之 1 I 則 檢 査 器 5 8 或 外 部 規 則 檢 查 介 面 40 » 輿 交 換裝 置 12 被 規 剌 注 意 事 i i 為 使 用 内 部 規 則 檢 査 器 58 或 外 部 規 則 檢 查 器42無 18 ηη 0 内 部 項 再 t 装 l 規 則 檢 查 器 58及 外 部 規 則 檢 査 器 42提 供 一 個決 策 決 定 埋 輯 >Ί 本 頁 t 來 決 一 指 定 資 料 封 包 的 目 的 MAC 埠 〇 因而決策機構送 1 I 出 —· 指 定 的 資 料 封 包 到 單 — 的 埠 多 埠 > 或所 有 的 埠 (亦 1 I 即 廣 播 )〇 例如, 每- -個資料封包的表頭都有來源及目的 1 i 訂 1 位 址 > 決 策 櫬 構 根 據 巨 的 位 址 辨 識 出 適 當 的輸 出 MAC埠 〇 另 一 方 面 該 0 的 位 址 相 關 於 —1 個 虛 擬 位 址(V i r t U e 1 1 a d d r e s S) 該 虛 擬 位 址 由 適 當 的 決 策 櫬 構 認為 相 關 於 複 數 1 I 個 m 站 〇 另 — 方 面 > 該 已 接 收 資 料 訊 框 根撺 IEEE802. Id [ 辣 1 1 協 定 將 包 括 VLAN (V i Γ t U e L AN ) 摞 m 訊 框 (t a g g e d f r am e ) 該 财 定 規 定 別 的 網 路 (g ί 1 00 M\bp s網站22群中的- -個路由 1 1 1 器 )及前述網站群的規則。 因此, 内郜規則檢査器58或經 1 1 I 介 面 40 的 外 部 規 則 檢 査 器 42 * 將 決 定 暫 時 貯存 在 媛 衝 記 憶 1 1 體 36的 訊 框 是 要 输 出 到 單 一 個 MAC埠或複數涸HAC埠 0 1 i 使 用 外 部 規 則 檢 查 器 42有 下 列 的 優 點 ,如 第 一 » 增 如 j 容 最 第 二 > 在 訊 框 未 完 全 送 到 外 部 鍰 衝 記憶 體 之 刖 1 產 1 1 生 --* 隨 意 基 礎 次 序 (Γ ad 〇 m -b as e d 0 r d e r in g ), 該 随 意 基 礎 1 I 次 序 係 位 於 致 能 訊 框 轉 送 決 策 的 決 策 佇 列 (dec is i 〇 η 1 1 本紙張尺度適用中國國家標丰(CNS ) Λ4規格(2丨OX 297公釐) 9 1472 4 53 Α7 Β7 經"部中央標??局貝1消費合作社印餐 五、發明説明(10) q u e u e )中;第三,致能淹生一次序中的決策,而與交換裝 置丨2中已接收訊框中的次序無關。 該決策機構(即内部規則檢查器5 8或外部規則檢杳器 42), Μ増向最(port vector)的形式輸出一個轉送決策到 開闞次系統5 6 ,其中埠向最可識別每一應接收資料封包的 MAC埠。由規則檢查器來的埠向量中有位址資料,瑄些位 址位於外部記憶體36中並貯存著資料封包。埠向量中堪有 用來接收資料封包Μ供傳輸的MAC埠的識別資料(如MAC埠 0 - 2 6 ) P交換裝置次系統5 6 ,經由外部記憶體介面3 4 ,從 外部記憶體3 6提取經埠向虽識別後之資料封包,提供此重 新取回的封包(ret rived data packet)到適當的、已識別 埠的傳送FIFO 54。 更多的介面提供管理及控制資訊。例如,管理資料介 面59根撺MII管理規格(IEEE 802, 3u),致能交換裝置12, 使它跟交換裝置送收器16、 100 M\bPs實體裝置20交換控 制及狀態資訊。管理資料介面59輸出一管理資料時 (MDC ),供姶雙向管理資料10信號路徑做為其時序基準。 PCI介面26是修正2.1版相容僕界面(slave interface >的3 2位元P C I介而,提供P C I主機處理器到内部I H S狀態及 配置暫存器60的資料存取,及提供外部記憶體36的資料存 取。若在複數交換裝置裝置的場合,PCI介面26也可做為 一個擴充R流排。管理埠30介面經一標準7線(seven wi「ed>反串列GPSI介面到一外部MAC機構,致能一主機控 制器,該主機控制器係以摺準MAC麿協定來存取交換裝置 本紙張尺度適用中國國家標準(CNS ) Λ4規格(210X 297公釐) ,Λ „ , ,,„ 10 91472 --τ----I--‘裝------訂------地 (請先閱讀背面之注意事項再填'"本頁) 經濟部中央標If局Λκ消费合作社印" 4 幻。., A7 B7 五、發明説明(1 1 ) 1 2 〇 第3關是根撺本發明之示範宵腌例,對第2圖交換裝 置次系妹56更詳细之敘述。第2圖的交換裝置12的其它元 件在销3圓中畢書,Μ說明交換裝置次系統5 6跟這呰原件 的關係。交換裝貿次系統56具有一個核心交換機構(core switchiriK engine),用來接收和轉送訊框。該用來完成 交換機構的主功能性方塊圖包括:一個埠向量FIF070, — 個媛衝管理.員(b u f f e r m a n a g e r ) 7 2 ,複數個埠輸出佇列( port output queue)74, — 個管理増輸出 f?列( management port output queue)75, —傾搞充匯流排璋 輸出庁列(expansion bus port output queue)77, —個 空閒媛衝槽(free buffer poo丨)104, —多複製佇列( multiple copy queue)90,多複製快取(multiplecopy cache)95, —回收庁列(reclaim queue)98。這些功能性 方塊的操作及構造將會更詳细地說明,但第3圖交換裝置 次系統5 6的簡單槪述首先提出,Μ便帶出下面的討論。 由埠進人多埠交換裝置12有兩個基本型態的訊框:單 複製 Un icopy)訊框及多複製Uu 1 t〖copy)訊框。單複製 訊框是一訊框由一個埠接收,並將由多埠交換裝置12傳送 至另一埠。相對的,多複製訊框是一訊框由一個埠接收, 並將送至超過一個Μ上輸出埠。第三圖中每一埠可由一個 MAC埠50代表,MAC埠50並有自己的接收FIFO 52,及傳輸 FIF0 5 4 〇 不管單一的訊框或多複製訊框都是由内部的HAC機 本紙乐尺度適用中國國家標準(C'NS) A4規格(210X 297公釐) ,„ η-, Λ η η 1 1 9 14 7 2 ^ ' -裝 訂 線 (請先間讀.背而之注意事項再填8本I〕 Ο53 () S Λ7 B7 _ V. Description of the invention (6) Send, receive the carrier's de-acknowledgement on the sensing medium, and wait for the packet gap period, it will try to send precious materials, and it will cause collision. Therefore, the website transmitting the data will monitor that the medium is a collision between the two websites to be transmitted at the same time. If a collision is detected, the two websites will stop transmitting data, wait for a random time, and then restart. Try transmitting. The website of 100M / bps22, according to the Ethernet standard IEEE802.3X full-flow control full-duplex working document (0.3), suggests that the full-duplex chess operation will be better. A full-duplex environment provides a two-way, point-to-point communication path between each 100 M / bps network website 22 and a switching device 12, where the switching device 12 and the website 22 can simultaneously transmit, Receive data without collision. The 100M \ bps websites 22 each have a 100H \ bps physical device (PHY) 208S of 100Base-TX, 100Base-T4, and 100Base-FX connected to the network medium 17. The switching device 12 includes a medium-independent interface (Mil) 24, which is connected to the physical device 20. The 100M \ bps website 22 can be operated as a server or router to connect to other networks. The central label of the Ministry of Foreign Affairs ^^ Pei Gong Consumer Cooperatives ^ (Please read the notes on the back before filling this page) As shown in Figure 1, the network 10 includes a series of transceivers (transceiver) 16 Use time division ninlUpiexing and time division demultiplexing to transmit 14 data packets transmitted between the switching device 12 and the 10M \ bps website. A magnetic transformer chess set 19 maintains the signal waveform on the medium 17. The switching device 12 includes a transmitter-receiver interface 18 for sending and receiving data packets from the switching device transmitter-receiver 16 and serially connecting N (? Z (series non-return to zero) The interface 23 uses a time division multiplexing communication protocol to accomplish this task. The exchange device receiver 16 receives packets from the HRZ 23 in series, and the paper size applies the Zhongyuan National Standard (CNS) Λ4. 210X297 public address) 6 91472 4 53 〇A7 B7 4 £ The cooperation between the laborers of the Central Bureau of Standards and the Ministry of Printing, printed by the company V. Description of the invention (7) Multiplex the received packet, and pass the network medium 1 7 Output the packet to the appropriate terminal website 14. According to the practical example of this disclosure, each exchange device receiver 16 has four independent twistedpair ports, and uses 4M multiples via the NRZ interface. Therefore, the overall pins are four times less than those directly connected by the switching device 12. The switching device 12 has a decision making engine, a switching engine, and a buffer memory interface. Configuration / control / status register, management count (Management counter) R Μ Λ C (in edia access control) protocol interface interface is used to support the routing of data packets between Ethereum websites 14, 22. Switching device 12 is also capable of intelligent switching decisions. Enhance the function and provide the HIB management information base (HIB management information base) object type statistics. The network information is provided to the external management entity, which will be described later. The switching device 12 also has an interface for enabling external data packets. Storage and switching Ugic M reduce the chip size of the switching device 12. For example, the switching device 12 has a synchronous dynamic RAM (SDRAM) interface 34 for external memory 36 to access, and external memory 36 is used to store Received frame, memory structure, and MIB counter information. The external memory 36 can be 80, 100, 1 20HHZW SDRAM with 2 to 4Mb memory capacity. The switching device 12 also includes a management port 30, which is a management port. 30 Enabling the external management body enables the entity to control the overall operation of the switching device 12 by managing the MAC interface 3 2 (magement MAC inter · fae 32). Set 1 2 ft packet and a PCI interface 2 (5, used to enable the host and bridge via PCI (Notes for reading before and reiterated on this page) This paper standard applies to China's standard (CNS) Λ4 specification (210XW7 male foot) 7 91472 * 53 os. Printed by the Central Bureau of Standards of the People's Republic of China on the consumption reduction cooperation of 1 ;! Λ7 B7 V. Description of the invention (8) External management of the external access of the device 2 8 On the other hand, if the PCI host and the bridge 2 8 are plural In the case of a single exchange device 12, the n exchange device 12 will be used as if it is a full stream. The exchange device 12 includes an internal decision-making mechanism (second picture), which internally decides to selectively transmit data from a data source. Information packets to at least one destination website. The internal decision-making body can also be replaced by an external rule checker ("u 1 echecker"). The switching device 12 includes a rule checker interface (ERIC) 40, allowing the external rule checker to generate a frame transfer Decision (frame forwarding dec 〖sion) M replaces the internal decision-making mechanism. Therefore, the frame forwarding decision can be generated by the internal exchange mechanism or by the rule inspector. The switching device 12 also includes an LED interface 44 for outputting the status of each port. And actuating the LED external winding series 46, which in turn actuates the human-readable LED element 48. A stack of oscillators 38 provides a 40MHZ clock wheel to the switching device 12. 2nd is _ Square diagram of the overall multi-port switching device 12. The switching device 12 includes 24 10M \ bps media access control ports (50), which are used at the respective 10M \ bps websites (増 1-24 ), M data packets are transmitted and received in half-duplex mode. Two 100M \ bps HAC ports 53 are used for each 100M \ bps website (埤 25, 26). Way to send and receive data packets. The management interface 30 also operates according to the MAC protocol (port 0). Each of MAC ports 50, 53, and 30 has a receive first-in-first-out (FIFO) buffer 52 and transmits FIF054, a data packet from the website Received by the corresponding MAC port, and stored in the phase I 1 ^^^ * ^^^^ 1 ^^^^ 1 — ^ n tLH -J. ) This paper size applies Chinese National Standards (CNS) Λ4 specifications (210X 297 feet) 8 91472 4 53〇A7 B7 ^:-Ministry of Standards and Construction Cooperative Consumers ’Printing Co., Ltd. V. Invention Description (9) 1 Corresponding FI F0 52 0 The received data packet is output from the corresponding F IF0 52 to the external J · t an memory interface 34 so as to be stored in the external memory 36 0 1 1 header of the received packet At the same time, it is sent to the decision-making body «Whether 1 1 is the internal rule checker 58 or the external rule check interface 40 first, Μ decides which I will read the data from the AC port. In particular, the packet header is sent to the internal rule 1 I checker 5 8 or external rule checker interface 40 »public exchange device 12 is regulated notice ii is to use internal rule checker 58 or external rule checker 42 None 18 ηη 0 The internal term is reinstalled. The rule checker 58 and the external rule checker 42 provide a decision-making record. This page t determines the destination MAC port of the specified data packet. Therefore, the decision-making agency sends 1 I out. — · Specified data packets to a single port—multi-ports> or all ports (also 1 I is broadcast). For example, the header of each data packet has a source and destination 1 i Order 1 address > The decision-making structure identifies the appropriate output MAC port based on the giant address. On the other hand, the address of 0 is related to-1 virtual address (Virt U e 1 1 addres S). The virtual address is determined by the appropriate address. Decision structure is considered to be related to plural 1 I m station 〇 Another — aspect> The received data frame is based on IEEE802. Id [The 1 1 protocol will include VLAN (V i Γ t U e L AN) 摞 m frame (taggedfr am e) The financial regulation The rules of other networks (g 1 00 M \ bp s website 22 group-a router 1 1 1 device) and the aforementioned website group. Therefore, the internal rule checker 58 or the external rule checker 42 via the 1 1 I interface 40 * will determine whether the frames temporarily stored in Yuan Chong memory 1 1 body 36 are to be output to a single MAC port or a plurality of HAC ports. 0 1 i The use of the external rule checker 42 has the following advantages, such as the first »Incremental j The second content> The frame is not completely sent to the external buffer memory 1 Product 1 1 Product-* Random The basic order (Γ ad 〇m -b as ed 0 rder in g), the random basis 1 I order is located in the decision queue of the decision-making for enabling frame transfer (dec is i 〇η 1 1 This paper standard applies to Chinese national standards Fung (CNS) Λ4 specification (2 丨 OX 297 mm) 9 1472 4 53 Α7 Β7 The Ministry of Central Standardization Bureau Bei 1 Consumer Cooperatives printed meals 5. Description of the invention (10) queue); Decisions in a sequence can be overwhelmed regardless of the sequence of received frames in the switching device 2. The decision-making body (ie, the internal rule checker 58 or the external rule checker 42), the MH outputs a transfer decision to the port system 5 6 in the form of a port vector, where the port direction can most identify each The MAC port that should receive data packets. The port vector from the rule checker contains address data, some of which are located in external memory 36 and store data packets. The port vector can contain the identification data of the MAC port used to receive the data packet M for transmission (such as MAC ports 0-2 6) P switching device sub-system 5 6, extracted from the external memory 3 6 through the external memory interface 3 4 Through the port, although the identified data packet is provided, the retrieved packet (ret rived data packet) is transmitted to the appropriate, identified port's transmit FIFO 54. More interfaces provide management and control information. For example, the management data interface 59 is based on the MII management specification (IEEE 802, 3u) and enables the switching device 12 to exchange control and status information with the switching device receiver 16, 100 M \ bPs entity device 20. When the management data interface 59 outputs a management data (MDC), the two-way management data 10 signal path is used as its timing reference. The PCI interface 26 is a 32-bit PCI interface that revises the 2.1 compatible slave interface (slave interface >), and provides data access from the PCI host processor to the internal IHS status and configuration register 60, and provides external memory. 36 data access. In the case of multiple switching devices, the PCI interface 26 can also be used as an extended R stream. The management port 30 interface is connected to an external via a standard 7-wire (seven wi "ed > anti-serial GPSI interface) The MAC mechanism enables a host controller that uses the standard MAC 麿 protocol to access the exchange device. The paper size applies the Chinese National Standard (CNS) Λ4 specification (210X 297 mm), Λ „, ,, „10 91472 --τ ---- I-'installation ------ order ------ land (please read the precautions on the back before filling in' " this page) Ministry of Economy Central Standard If Bureau ΛκConsumer Cooperative Association's "4 .., A7 B7 V. Description of the invention (1 1) 1 2 〇 The third level is a demonstration of the example of the present invention, for the second picture of the exchange device sub-line 56 more Detailed description. The other components of the exchange device 12 in FIG.呰 The relationship between the originals. The exchange loading and exporting sub-system 56 has a core switchiriK engine, which is used to receive and forward message frames. The main functional block diagram used to complete the exchange mechanism includes: a port vector FIF070,- Geyuan Chong Management. Staff (buffermanager) 7 2, multiple port output queues 74, — management port output queue 75 (management port output queue) 75, — charge the bus line output 充Column (expansion bus port output queue) 77, — a free buffer poo (104) — — multiple copy queue (90), multiple copy cache (95), — recycling queue (Reclaim queue) 98. The operation and structure of these functional blocks will be described in more detail, but the simple description of the switching system sub-system 56 in Figure 3 is first proposed, and M brings out the following discussion. The human multi-port switching device 12 has two basic types of frames: a single copy Un icopy) frame and a multiple copy Uu 1 t 〖copy) frame. A single copy frame is a frame received by a port and will be received by a port. Multi-port switching 12 transferred to another opposite port. In contrast, a multi-copy frame is one frame received by one port and sent to more than one M output port. Each port in the third figure can be represented by a MAC port 50. The MAC port 50 has its own receiving FIFO 52, and the transmission FIF 0 5 4 〇 No matter a single frame or multiple copy frames are made by the internal HAC paper music The dimensions are applicable to China National Standard (C'NS) A4 specifications (210X 297 mm), „η-, Λ η η 1 1 9 14 7 2 ^ '-Gutter (please read it first. Please fill in the precautions before you fill in 8 books I] 〇

G Α7 Β7 經濟部中央標i?-局Α工消費合竹社印絮 五、發明説明 ( 1 2 ) 1 I 構 5 0接 收 〇 當 訊 框 封 包 由 埠 接 收 時 4 該 封 包 被 置 於 接 收 I F ί F 0 5 2 〇 每- -封包有- -個表頭, 該表頭用來提供給規則 1 1 檢 杳 器 t 不 管 是 內 部 規 則 檢 查 器 58或 外 部 規 則 檢 查 器 42皆 1 ffl 先 可 内 郜 規 則 檢 查 器 5 8 或 外 部 規 則 檢 查 器 4 2 t 根 據 表 頭的 閱 讀 資 訊 決 定 該 訊 框 封 包 要 投 往 何 處 9 即 訊 框 封 包 要 傅 送至〖J 背 1¾ 1 I 之 1 注 i 哪 一 個 埠 或 哪 些 埠 ? 意 \ 事 1 I 内 部 規 則 檢 查 器 5 8 或 外 部 規 刖 撿 查 器 42 在 m 造 其 轉送 Jn 再 1 裝 [ 決 策 的 時 候 > m 衝 器 管 理 員 7 2 由 空 閒 緩 衝 槽 10 4接收到一 % 本 頁 個 空 閭 m 衝 指 摞 ( f Γ e e b u f f e Γ P 0 in t e r ), 此 空 間 媛 衝指 I I 標 是 外 部 記 憶 體 36的 一 個 位 址 » 接 收 FI F0 5 2會將訊框儲 1 1 | 存 在 該 處 〇 一 旦 緩 衝 器 管 理 員 由 空 閒 緩 衝 槽 10 4接收到空 1 1 訂 閒 媛 衝 指 標 , 由 空 間 緩 衝 指 標 指 到 (P o i n t )的鍰衝器將不 1 再 被 認 為 是 空 閒 的 > 該 訊 框 資 料 Μ 直 接 記 憶 體 存 取 (DMA) 1 I 的 執 行 方 式 ♦ 由 接 收 FIFO 5 2經資料匯流排8 0 , 傳送到外 1 1 部 記 憶 體 3 6 〇 雖 然 有 些 m 衝 器 會 儲 存 訊 框 1 此 將 後 述 ,在 Ϊ 此 該 訊 框 資 料 是 存 於 由 空 閒 緩 衝 槽 10 4來的空閒緩衝指標 1 I 所 指 到 的 位 址 〇 1 1 [ 除 了 表 頭 資 訊 之 外 , 外 部 規 則 檢 查 器 42或 内 部 規 則檢 1 1 査 器 58 也 從 緩 衝 器 管 理 員 7 2 9 接 收 到 一 個 空 閒 緩 衝 指標 1 1 , 此 空 閒 媛 ftB-r 衝 指 標 琨 在 被 稱 為 訊 框 指 標 (f r a me P 0 1 n t e r ) ί > 因 為 它 指 到 存 有 訊 框 的 外 部 記 憶 體 36 内 的 位 址 〇 外 部規 J 則 檢 查 器 4 2 * 或 内 部 規 則 檢 査 器 5 8 使 用 表 頭 的 資 訊 製 造轉 1 1 送 決 策 r 並 Μ 埠 向 虽 的 形 式 產 生 轉 送 指 令 ύ 在 本 示 範 說明 1 1 I 實 施 例 中 埠 向 最 是 28 位 元 的 向 最 9 有 一 組 位 元 組 (b it 1 1 12 本紙張尺度適用中®國家標苹(CNS ) Λ4規格(2I0X297公釐) 91472 經满部中央標準局負工消費合作社印繁 4 A7 in 五、發明説明(13) s e t )供每一輸出埠傳送訊框到該訊框該送到的地方。假設 存:本概述例中,已接收訊框是一單複製訊框,由外部規則 檢杳器4 2或内部規則檢査器5 8產生的埠向畐,只有一個位 元被設定。被設定的位元係有關於埠群中的某一個埠。 外部規則檢查器42或内部規則檢査器58放一個埠向量 及訊框指標(遇有控制操作碼(control opcode)及VLAN標 記(Udex)>到埠相最FIFO 70。埠相量FIFO 70檢查埼向量 Μ決定該跟埠相量有關的訊框指標要放入哪一個輸出貯列 7 4 (或是哪些個複數個輪出佇列7 4 ),。輸出佇列7 4將訊框 指標放在適當的輸出佇列7 4的頂端。如此會使訊框的傅送 排成一列。 在時間的某一點,該訊框指標穿過輸出佇列74到其底 端。當訊框指標到輸出佇列74的底端時,缓衝管理員提取 此訊框指搮,並經由訊框指標讀取匯流排(f「aroe pointer read bus)86,傳遞該訊框指標到適當埠的傳送FIFO 51 如此會排程(schedule)訊框傳輪的時間,訊框資料將由訊 框指標指到位於外部記憶體36内的一個位址,以DHA的方 式被讀取,目前則置於適當的FIFO 54,等待傳送。多複 製傳輸跟單複製傳輸類似,除了埠相量中多個位元被設定 外。這些位元被設定之後,將可指定訊框要傳送到哪些涸 埠。訊框指標被放人適當的輸出佇列74,並由相闞的傳輸 FIF0 54傳出。 緩衝器管理員7 2使用特殊的控制佇列,即空閒鍰衝槽 (free buffer pool)104,多複製庁列(multiple copy 本紙張尺度適用中國國家標準(CNS ) A4規格(210X 297公釐) 13 91472 C誚先閲漬背而之注意事項再峨巧本頁) 裝_G Α7 Β7 Central Standard of the Ministry of Economic Affairs-Bureau A Industrial Consumers Co., Ltd. Printing V. Description of the Invention (1 2) 1 I Structure 5 0 Reception 〇 When the frame packet is received by the port 4 The packet is placed in the receiving IF ί F 0 5 2 〇 Each--packet has--a header, which is used to provide the rule 1 1 checker t whether it is the internal rule checker 58 or the external rule checker 42 1 ffl can be inside郜 Rule checker 5 8 or external rule checker 4 2 t Decide where to send the frame packet 9 according to the reading information of the header 9 That is, the frame packet is to be sent to J Back 1¾ 1 I 1 Note i Where One port or which ports? Meaning \ Event 1 I Internal rule checker 5 8 or External rule checker 42 Make it forward Jn and then install it [When making a decision> m Driver manager 7 2 Buffered by free Slot 10 4 Received 1% of this page ’s empty frame (f Γ eebuffe Γ P 0 in ter). This space is labeled II as an address of external memory 36 »Receive FI F0 5 2 will send a frame Store 1 1 | There is this place. Once the buffer manager receives the empty 1 1 from the free buffer slot 10 4 and subscribes to the free punch indicator, the buffer pointed to by the spatial buffer indicator (P oint) will not be reset by 1 again. It is considered to be idle > The frame data M Direct memory access (DMA) 1 I implementation method ♦ Received FIFO 5 2 through data bus 80 0, transferred to external 1 1 memory 3 6 〇 Although some The m buffer will store frame 1, which will be described later. Here, the frame data is stored in the address indicated by the free buffer index 1 I from the free buffer slot 104. [1 In addition to the header information, Outside, external rule checker 42 or internal rule check 1 1 check 58 also received a free buffer indicator 1 1 from the buffer manager 7 2 9. The free buffer ftB-r punch indicator is called a frame indicator (fra me P 0 1 nter) because it refers to The address in the external memory 36 where the frame is stored. 0 The external rule J rule checker 4 2 * or the internal rule checker 5 8 uses the information of the header to make a turn 1 1 sends a decision r and generates a port direction. Forwarding instructions In this 1 1 I example, the port direction is the most 28-bit direction and the 9th direction has a set of bytes (b it 1 1 12 This paper is applicable to the National Standard Apple (CNS) Λ4 specification ( 2I0X297 mm) 91472 Printed by the Central Bureau of Standards and Consumers' Cooperatives 4 A7 in V. Invention Description (13) set) For each output port to send a frame to where the frame should be sent. Assumption: In this overview example, the received frame is a single copy frame. Only one bit is set for the port direction frame generated by the external rule checker 42 or the internal rule checker 58. The set bit is related to a certain port in the port group. The external rule checker 42 or internal rule checker 58 puts a port vector and frame index (in case of control opcode and VLAN tag (Udex)> to the port phase FIFO 70. The port phasor FIFO 70 check The unitary vector M determines which output bank 7 4 (or which multiple round-out queues 7 4) the frame indicator related to the port phasor is put into. The output queue 7 4 puts the frame indicator At the top of the appropriate output queue 7 4. This will make the frame's Fu send into a line. At some point in time, the frame indicator passes through the output queue 74 to its bottom. When the frame indicator reaches the output At the bottom of queue 74, the buffer manager extracts this frame pointer, and reads the frame indicator f (aroe pointer read bus) 86, and passes the frame indicator to the appropriate port transmission FIFO 51. This will schedule the timing of the frame transfer. The frame data will be pointed from the frame pointer to an address located in the external memory 36, read in DHA mode, and currently placed in the appropriate FIFO 54. , Waiting for transmission. Multi-copy transmission is similar to single-copy transmission, except that there are many The individual bits are set. After these bits are set, it can specify which port the frame is to be transmitted to. The frame index is put into the appropriate output queue 74 and transmitted by the corresponding transmission FIF0 54. The buffer manager 7 2 uses a special control queue, namely the free buffer pool 104, multiple copy queue (multiple copy) This paper size applies to China National Standard (CNS) A4 specifications (210X 297 mm) 13 91472 C 诮 Please read the precautions before you read this page before installing this page)

'1T i-v 4^3〇8 Μ Β7 五、發明説明(14) qneUe)90.回收佇列9 8,及多複製快取96。這些持殊的控 制貯列,會管理配置緩衝器的程序,Μ儲存已接收訊框。 及管理恢復的緩衝器(retriving buffer), Μ便一旦這 呰訊框送到它被指定該送到的埠後,重新使用這些緩衝器 。鍰衝器管理R 7 2也為輸出佇列7 4 ,控制佇列1 0 4、9 0、 98維請(maintain)位於外部記憶體36的溢位區(overflow region),此一部份容後詳述。 以上述操作概述作為背景,交換裝置次系統56的每一 傾部份,及各種觀點在此將更詳细地討論。本發明的第一 個觀點是,本發明中有不同的輸出佇列74的結構。輸出疗 列74指定供給10 Mb/s輸出埠及100 Mb/s輸出埠使用外, 尚有輸出佇列75供管理埠30使用,輸出佇列77供掮充埠 26使用。這些輸出佇列75、 77跟輸出佇列74有同樣的外部 (請先閱讀背而之注意事項再填寫本頁) 後 於 述 软 將 此 置 配 部 内 的 同 不 有 但 構 結 結 部 外 勺 Ad 4 7 歹 佇 出 輸 例 胞 茛 1 明 發 本 撺 根 是 圖 4 第 部 個三 有 4 7 列 佇 出 輸 出 看 地 顯 明 很 圖 四 第 由 C 圖 塊 方 構 结 列 序 的 有 所 片 晶 該 持 保 好 最 •1IJ1 有 性 的 高 最 有 要 果 如 經济部中央標準局月工消f合作社印絜 置 裝 換交 埠 多 照 參 /1" 構 高 極 將 值 價列 的排 片要 晶 卻 該 , , 用 話 使 的置 做裝 樣換 這 交 果 個 如 一 C- 給 ) 十 2-1Ε 1Ϊ 設 片 晶 該 當 一 區 供量 提容 由 憶 經 記 明低 發 能 本性 〇 高 題有 問具 的上 難片 兩 晶 個在 一 列 為 作 成 該 此 , . 列 人庁 輸出 的輸 董單 大個 當該 0 然 題雖 問 , 的列 難佇 兩 的 一 最 此 容 決大 解 個 此 一 Μ任 , 擔 區列 載佇 過許 一 允 有區 外載 片過 晶 , 在時 而要 , 需 低 要 能 性 域 區 的 上 Η 晶在 It 區 91472 本紙張尺度適用中國國家標準(CNS ) Λ4規格(210X297公釐) 453〇 A7 B7 經滴部中央標準局員工消费合作社印掣 五、發明説明( 15 ) 1 | 根 撺 第 4圖的實施例. 本發明的單理輯輪出佇列74, I 有 個 莨 體 區 〇 此 包 括 一 個 輸 出 佇 列 書 寫 端 76 t 輸 出 貯 列 1 .I 讀 取 端 78 » 及 位 於 外 部 記 憶 體 的 輸 出 fr 列 γη 過 載 區 (- -般指 讀 閱 1 定 為 11 0 ) η 如 上 所 述 所 有 的 輸 出 佇 列 7 4都經 由 外 郜 記 憶 Ί m 界 面 3 4存 取 外 部 記 憶 體 36 〇 本 發 明 利 用 本 外 部 記 憶 體 的 背 面 之 I i 組 發 本 質 (b U Γ S t in n a t U re)的優點, 使得該資料(訊 框 指 注 意 事 ί t 1 標 ). 經由連接晶片1 2及外部記憶體3 6的匯流排8 4 , 送進 J貝 再 填 1 裝 I 及 送 出 過 載 區 11 0 〇 η 本 頁 輸 出 佇 列 書 寫 端 76及 輪 出 佇 列 謓 取 端 78位 於 晶 片 12之 1 I 上 Ο 輸 出 佇 列 書 寫 端 76及 輸 出 佇 列 謓 取 端 78被 認 為 是 小 且 1 1 I B 貴 的 資 湄 0 相 反 的 1 組 成 輸 出 佇 列 74的 第 三 部 分 « 過 載 1 1 區 1 1 0則被認為是大且便宜的, 當通往過載區的路徑提供 訂 1 一 個 低 性 能 高 容 量 的 路 徑 時 t 輸 出 佇 列 書 寫 端 76及輸 出 1 1 佇 列 m 取 端 78題 , 提 供 出 高 的 性 能 0 1 i 操 作 上 1 輸 出 佇 列 書 寫 端 76接 收 一 個 輸 入 資 料 Ο 根 據 1 冰 1 I 本 發 明 多 埠 交 換 裝 置 1 2的 示 範 實 胞 例 該 輸 入 資 料 是 一 個 訊 框 指 標 * 用 來 指 到 外 部 記 憶 髖 的 第 一 個 m 衝 器 9 該 缓 衝 1 1 I 器 存 有 個 訊 框 的 刖 2 5 6位元組。 這些技转的技巧是很容 1 1 易 明 牌 的 然 而 1 不 管 是 本 多 埠 交 換 裝 置 1 2或 是 別 的 技 術 1 1 〇 輸 出 佇 列 結 構 體 74並 沒 有 限 制 其 輸 入 資 料 的 型 態 只 有 訊 ;1 Ί 框 指 標 —’ 種 ψ 而 是 可 廣 泛 地 排 列 其 它 形 式 的 輸 入 資 料 〇 在 輸 入 資 料 通 過 輸 出 佇 列 書 寫 端 76並 到 達 其 底 端 之 後 1 1 » 跟 輸 出 佇 列 74有 闞 的 控 制 埵 輯 決 定 該 如 何 處 置 該 輸 人 資 1 I 料 Λ 如 果 在 輸 出 fr 列 講 取 端 78尚 有 空 間 9 輸 出 佇 列74的 過 1 1 15 本紙張尺度適中國國家標準(CNS } Λ4規格(2iOx297公釐) 9 1472 Ο c Α7 Β7 五、發明説明(16 的 空 是 ο 1 1 區 載 軎 歹 貯 出 輸 由 將 料 資 入 輸 的 多 更 或 6 7 端 寫 書 列 fr 出 輸 由 此 ο 8 7 端 取 讀 列 rf 出 輸 到0 傳 6 7 端 完 作 的 8 7 端 取 讀 列 佇 出 輸 到 料 資 入 輸 的 多 更 或 1 遞 傳 間 時 待 等 低 個 一 ο 是動 此流 因料 。 資 的人 行輸 執 的 上 速 ]2快 片 , 晶 Y. 在nc 是te 全la 端 寫 書 nu 歹 佇 出 輸 且 並 滿 已 8 7 端 取 讓 列 佇 出 輸 果 如 寸 尺 發 個 一 有 少 至 大 元 位 6 1X 即量 料 資 的 區 載 遇 4 7 列 佇 出 輸 到 寫 式 形 的 發 組 Μ 料 資 人 輸 該 則 小 端 寫 書 列 ff 出 輸 在 但 滿 已 8 7 端 取 讀 列 佇 出 輸 果 如 寸 尺 發 組 個 一 有 沒 並'1T i-v 4 ^ 3〇8 Β7 V. Description of the invention (14) qneUe) 90. Recovery queue 9 8, and multi-copy cache 96. These special control banks will manage the buffer allocation process, and M stores the received frames. And manage the retriving buffers, once the frame is sent to the port to which it is assigned, the buffer is reused. The buffer management R 7 2 is also the output queue 7 4, and the control queue 1 0 4, 9 0, 98 is maintained in the overflow region of the external memory 36. This part of the content Details later. With the above operation overview as background, each aspect of the switching device sub-system 56 and various viewpoints will be discussed in more detail here. A first aspect of the present invention is that there are different configurations of the output queue 74 in the present invention. The output therapy line 74 is designated for 10 Mb / s output port and 100 Mb / s output port. There is also an output queue 75 for the management port 30 and an output queue 77 for the charging port 26. These output queues 75, 77 have the same external appearance as output queue 74 (please read the precautions before filling this page), and then describe the soft parts in the disposition section, but the external structure of the structure section. The example of Ad 4 7 was lost. Cell buttercup 1 The root of the Mingfa is shown in Figure 4. There are 4 7 columns in the output. The output is obvious. Figure 4 shows the sequence of the C block. The most important thing is to maintain the best quality of 1IJ1. It is necessary to have a high sex ratio. For example, the Central Standards Bureau of the Ministry of Economic Affairs, the National Labor Bureau, the cooperatives, the cooperatives, the printing and installation of the port, and the replacement of the port. Jing should deserve to, use the words to make the sample change the result like a C-giving) Ten 2-1 Ε 1Ϊ Set the piece of crystal should be a district to increase the capacity by Yi Jing recorded the low energy nature There are two pieces of puzzles on the upper part of the questionnaire. This is done in a row. The list of losers from the list of people is a big one. When the question is asked, one of the two is the most important solution. This task许 Xu Yiyun has crystals outside the zone. Sometimes, it is necessary to have a low-performance region on the top of the crystal. In the It area 91472, this paper size applies the Chinese National Standard (CNS) Λ4 specification (210X297 mm) 453. A7 B7 Printed by the Consumer Standards Cooperative of the Central Bureau of Standards of the Ministry of Education 5. Description of the Invention (15) 1 | According to the embodiment of FIG. 4. The single series of the present invention is 74, I has a carcass area. This includes an output queue writing terminal 76 t output bank 1. I reading terminal 78 »and an output fr column γη located in external memory γη overload area (--generally refers to reading 1 as 11 0) η as described above All output queues 7 4 are accessed through external memory, m interface 3 4 accesses external memory 36. The present invention utilizes the nature of the I i group on the back of the external memory (b U Γ S t in nat U re) The advantage is that the information (the frame refers to the attention matter t 1 mark). It is sent to the J Beyer via the bus 8 4 connecting the chip 12 and the external memory 36. Fill 1 Load I and send the overload area 11 0 〇η The output queue writing end 76 and wheel output queue taking end 78 on this page are located on 1 of the chip 12 I on the output queue writing end 76 and output queue taking end 78 is considered to be small and 1 1 IB is expensive 0 Opposite 1 constitutes the third part of output queue 74 «Overload 1 1 Zone 1 1 0 is considered to be large and cheap when the path to the overload zone To provide a low-performance, high-capacity path, t output queue writing terminal 76 and output 1 1 queue m take 78 questions, provide high performance 0 1 i operation on 1 output queue writing terminal 76 receives an input Data 0 According to 1 Ice 1 I The demonstration cell example of the multi-port switching device 12 of the present invention. The input data is a frame indicator * used to refer to the first m punch of the external memory hip 9 the buffer 1 1 I The device stores a frame of 2 5 6 bytes. These technical transfer techniques are very easy to understand. 1 1 Yiming brand. However, 1 whether it is the multi-port switching device 12 or other technology 1 1. The output queue structure 74 does not limit the type of its input data. ; 1 Ί frame index-'kind of ψ, but can be widely arranged in other forms of input data 0 after the input data passes the output queue writing end 76 and reaches its bottom 1 1 »There is 阚 control with the output queue 74 74 I decide how to deal with the input of human resources 1 I Material Λ If there is still space in the output terminal 78 of the output 9 9 of the output queue 74 1 1 15 This paper is in accordance with the Chinese national standard (CNS) Λ4 specification (2iOx297) (Centi) 9 1472 〇 c Α7 Β7 V. Description of the invention (16 is empty ο 1 1 zone contains the storage and output by the input and output of more and more or 6 7 end of the book list fr output and output from this ο 8 7 fetches and reads the rf output to 0 and passes 6 6 7 8 finished fetches and reads the output and input to the material input Lose more or wait for the next one to pass. This is the reason for this flow. The speed of the bank ’s bankruptcy] 2 fast films, Jing Y. In nc is te full la end nu 歹 歹伫 Loss and lose and the full 7 8 ends to give up the row. The output of the loser is as large as 6 1X. The area of the material is 4 7 rows. The output is written to the writing form. The sending group M. The material person loses the little-end writing book and ff. The output is at but the full 7 8 ends have to read the list.

S Γ U 量 料 資 的 則 入 輸 該 —:------聚— 請先閱讀背面之注意事項再填寫本頁) 端 寫 書 列 佇 出 輸 在 留 料 資 輸 後 最 至 〇 作 不 也 麽 容 夠 足 有 8 7 端 取 讀 列 佇 出 輸 當 來 出 空 會 8 7 端 取 讀 列 庁 出 區 載 過 在 時 此 果 如虽 料 資 的 寸 尺 發 組 個 一 納 容 可最 料 資 個 I 有 區 載 過 自 來 個 送 將 料 資 發 組 個 訂 端 取 請 列 庁 出 輸 到 統 傳 如 上 致 大 8 7 端 取 謓 列 佇 出 輪 構 結 的 列 庁 出 輸 在 輸 ο 料 資 人 輸 取 提 地 個 一 接 個 一 它 為 因 作 操 地 般 列 佇 的 經潆部中央標隼局I工消f合作社印鉍 能 功 合 組 Μ 上 致 大 6 7 端 寫 書 列 佇 出S Γ U For materials and materials, enter this: —-------- Poly — Please read the precautions on the back before filling out this page.) Is it enough to have 8 7 end reading and reading out, when the output is empty, the 8 7 end reading and reading out, and the output area is contained in this time. The material I has a zone and contains a tap. The material is sent to the ordering party, please list it and output it to the general transmission. As above, the big 8 7 terminal picks up the list. ο The materials are collected one by one by the author. It is written by the Central Ministry of Economic Affairs of the Ministry of Economic Affairs, the Ministry of Economic Affairs, the Central Bureau of Industrial Standards, the Cooperative Society, and the Bismuth Energy Functional Group. Column out

C η U 體 憶 記 部 外 人 寫 便 Μ 料 資 發 組 個 一 成 料 資 合 組 作 列 一 佇 換 出 轉輸 明人 發 放 本料 此資 因入 〇 輸 6 i 3 個 件 i 一 單 個 V e η e V e 話 的 要0 果 如 的 St料 Ur資 ID ut 件允 Ml 5 事 7 發端 0 寫 個 書 一 列 為佇 成出 4)輸 區 載 過 的 6 3 晴 Ann 憶 記 部 外 到 發 組 後 然 積 累 區 載 過 供 提 是 不 而 區 存 儲 的 宜 便 個 供 提 段 時 的 擠 擁 料 資 在 本紙張尺度適用中國國家標卒(CNS ) A4規格(2!0X297公釐) 91472 A7 B7 經消部中央標準局負工消费合作社印鉍 五、發明説明( 17 ) 1 I 昂 貴 的 R 在 少 數 場 合 才 用 到 其 功 能 的 晶 片 〇 儘 管本 發 明 I 的 過 載 區 1 1 0是位Κ晶片外, 過載區11 〇的 存 取 9 是Μ 種 1 1. 有 效 率 的 方 式 在 间 一 時 間 組 發 個 數 目 的 位 元 組資 訊 〇 讀 先 閱 讀 1 此 對 比 於 傳 铳 佇 列 结 構 9 由 佇 列 中 一 次 只 寫 或 只 諝取 __. 個 1 輸 入 資 料 是 不 一 樣 的 〇 背 '由 之 注 意 事 1 1 在 操 作 的 期 間 1 如 果 有 大 最 的 資 料 到 達 輸 出 佇列 74 t 1 1 這 些 輸 入 資 料 放 進 過 載 區 U 0 Μ避免晶片上佇列7 8的過載 項 再 4 f 裝 I 0 因 此 , 使 用 本 發 明 的 庁 列 結 構 , 丟 掉 訊 框 的 動 作大 都 可 本 F 遛 免 〇 同 時 過 載 區 110的容量可Μ很容易地經由改變外 1 I 部 記 憶 體 36的 大 小 > 而 調 整 其 容 量 〇 再 者 4 個 別 特殊 大 小 1 1 的 過 載 區 1 1 0可歸劃用Μ定做佇列的大小, 而不須影饗輪 1 1 出 佇 列 74的 性 能 〇 訂' 1 典 型 上 * 佇 列 是 —~1 種 有 先 後 次 序 的 結 構 • 具 有先 人 先 1 1 出 的 構 造 0 在 有 型 式 的 佇 列 例 如 » 回 收 佇 列 98及 空 閲 1 j 媛 衝 槽 104. 輸人資料的先後次序並不重要。 如果有可跬 1 由 輸 寫 端 1 0 0直接送資料到謓取端1 0 2 , 本 發 明 允 許資 訊 由 1 I 瑄 條 路 傳 送 9 卽 旁 路 供 佇 列 用 的 過 載 區 〇 P 要 該 資訊 並 不 1 1 f 是 對 次 序 敏 感 的 (0 r d e r -S e η t ί ν e ), 即使該資訊在有關的 1 1 過 載 區 » 都 允 許 如 此 操 作 ο 例 如 , 回 收 區 的 m 衝 器不 是 對 1 I 次 序 敏 感 的 t 就 可 如 此 運 作 , 因 為 當 緩 衝 器 不 再 須要 儲 存 1 '1 該 訊 框 後 最 終 會 回 到 空 閒 媛 JStr 衝 槽 的 -.丨.\ 閲 表 列 (f r e e 1 i S t * 1 1 ), 所以任何在該鍰衝器中的次序都是可接受的。 因此, | 當 該 資 料 並 非 對 次 序 敏 感 > 若 有 回 收 佇 列 98寫 到 外部 記 憶 1 I 體 ϊ.Πί 過 載 區 1 1 0的動作, 為避免頻寬產生變化, R要請取端 1 1 17 本紙張尺度適用中國國家標準(CNS ) Λ4規格(210X 297公釐) 9 1472 453 A7 B7 經^部中史標^扃負工消費合作社印袈 五、發明説明( 1 8 ) 1 I 1 0 2有空位, 該資訊會由書寫端100直 接 傳 送 到 讀 取 端 102 1 | 〇 回 收 ίΐ 列 9 8是 __. 棟 佇 f *> 1 VIJ 的 例 子 該 種 的 佇 列 排 列 對 次 序 1 不 敏 感 的 η 料 <-> 在 許 多 應 用 中 ♦ 仍 有 許 多 對 次 序 不 敏 感 的 1 1 先 資 料 類 型 i 因 此 > 本 發 明 的 持 色 是 找 到 了 種 佇 列 的 應 用 間 讀 背 1 1 t 排 列 此 揷 較 不 樣 的 資 料 型 態 1¾ I 之 1 注 I 如 本 發 明 範 例 實 胞 例 第 1 _及第2圖 描 述 之 多 埠 交 換 裝 意 事 1 置 1 共 有 28個 輸 出 佇 列 (每 —* 涸 相 關 於 —. 個 輸 出 埠 ) 其 項 再 填 1 中 24個 供 10Hb/s 使 用 者 埠 使 用 » 2涸供100 用 者 埠 使 寫 本 頁 裝 | 用 t 一 個 供 管 理 埠 使 用 » — 個 供 擴 充 歷 流 排 埠 使 用 (? 當 訊 、· 1 I 框 指 標 排 成 一 列 供 傳 輸 用 時 * 輸 出 佇 列 74 、 7 5 77提 供 給 1 1 訊 框 指 標 一 個 暫 時 儲 存 的 地 方 0 此 即 先 進 先 出 的 (FIFO) 1 1 訂 方 式 提 取 先 刖 經 FIFO 70依據埠向里所指示, 將訊框指標 1 寫 至 不 同 的 輸 出 佇 列 74 7 5 > 77之 — 的 該 種 形 式 之 埠 向 量 1 1 I 〇 本 發 明 較 佳 的 實 拖 例 中 不 同 的 輸 出 庁 列 74 75 N 77 1 1 沐 包 含 下 列 許 多 個 或 全 部 的 領 域 單 複 製 位 元 訊 框 指 標 1 I 控 制 操 作 碼 或 控 制 信 號 9 及 VLAN (虛 擬 區 域 網 路 )標記, 1 1 I 單 複 製 位 元 搮 式 該 訊 框 » 使 得 該 訊 框 只 會 送 到 單 一 傾 輸 出 I 1 埠 〇 訊 框 指 標 指 到 外 部 記 憶 餺 36上 的 訊 框 〇 控 制 操 作 碼 識 1 1 別 跟 訊 框 有 關 的 特 殊 資 訊 (新得知的訊框) 0 控 制 信 號 由 控 ,j 1 制 操 作 碼 來 的 資 訊 1 用 以 指 示 這 iih 埠 在 傳 輸 訊 框 之 前 如 何 1 掌 握 訊 框 〇 VLAN標記 提 供 給 必 須 附 加 輸 出 訊 框 (如果镉要 1 I 的 話 >的V L A Ν標籤(t a g) 一 個 參 考 1 然 而 » 這 些 範 例 只 是 一 1 I 個 m 例 而 已 本 發 明 可 Μ 運 用 到 其 它 不 同 形 式 輸 出 佇 列 的 1 1 1 8 本紙張尺度適州中國國家標隼(CNS ) A4規格(210X297公釐) 91472 Ο c Α7 Β7 .¾¾-部中央標皁局男工消费合作社印黧 五、發明ΐ克明 ( 1 9 ) 領 域 〇 輸 出 fj 列 7 4的 第 —’ 種 型 式 個 的 內 部 结 構 ,1 0 M b / S 輸 出 佇 列 盡 於 第 5圖, 1 0 Mb/ s輸出佇列7 4持有將送到1 0Mb/ S埠的訊抿用輪入資料, 本範例說明莨施例中, 瑄呰佇列 的 輸 m 佇 列 書 寫 端 7 6 持 有 3 2 個 輸 入 資 料 » 而 這 佇列 的輸 出 佇 列 m 取 端 78 , 則 持 有 1 6 個 輸 入 資 料 〇 每 一 個 1 OMb/s輸 出 佇 列 7 4 的 輸 入 資 料 > 都 有 簞 複 製 位 元 及 訊 框 指 標。 在本 發 明 多 埠 交 換 裝 置 的 範 例 寊 胞 例 中 VLAN 標 籤 並 不脔 要, 因 為 沒 有 附 加 到 1 0 Mb/s 埠 的 V L A N m 條 〇 第 二 種 輸 出 佇 列 74的 -—" 個 範 例 實 m 例 之 内 部 結構 , 100Mb/ s_出佇列盡於第6 圖 t 100Mb/ S輸出佇列74持有將 送 到 1 00 Μ b / s埠的訊框用輸入資料, 這些佇列的輸出佇列 書 寫 端 7 6 持 有 6 4 個 輸 人 資 料 * 而 這 些 佇 列 的 輸 出 佇列 謓取 端 7 8 1 則 持 有 16 個 輸 入 資 料 0 每 一 假 m 入 資 料 包 括一 個 VLAN 標籤, -個部分控制操作碼, -個單複製位元, 一 個 訊 框 指 標 r> 外 部 記 憶 體 3 6 的 個 範 例 圖 賁 於 第 7圖, 外部記憶體 整 體 的 容 量 可 能 為 4Mb , 雖然在其它的實胞例中, 有不同 的 容 量 , 根 據 本 發 明 t 作 為 過 載 區 的 外 部 記 憶 體 36 , 允許 只 要 簡 單 地 改 變 其 外 部 記 憧 體 9 即 可 增 加 或 減 少 輸出 庁列 的 大 小 η 比 將 所 有 佇 列 结 構 部 做 在 一 個 晶 Η 上 * Μ致 所有 佇 列 容 最 由 該 片 晶 片 的 製 作 所 決 定 而 言 1 是 比 較 好的 0 為 了 滿 足 交 換 裝 置 12 的 儲 存 m 求 * 外 部 記 憶 體36 的一 個 範 m 實 拖 例 配 置 了 下 列 的 區 域 » 空 閒 媛 衝 槽 過 載區 1 20 讀 先 閱 讀 背 1¾ 之 注 項 再 填 :裝 η 19 本紙張尺度適用中國國家榇準(CNS ) A4規格(210X 297公釐) 9 1472 IT-C η U Outsiders write down the materials. The materials are sent to a group of 10%. The materials are grouped together. They are exchanged and transferred to the beneficiary. This material is entered because of the input. 6 i 3 pieces i a single V e η e V e If you want 0 If the St material Ur ID ID ut pieces Allow Ml 5 Event 7 Origin 0 Write a book as a list of successes 4) 6 in the input area After the formation, the accumulated crowded material in the area when the supply and withdrawal is over and the storage and storage area is convenient for the supply and withdrawal section is applicable to the Chinese National Standards (CNS) A4 specification (2! 0X297 mm) at this paper scale 91472 A7 B7 Printed by the Central Bureau of Standards of the Ministry of Consumer Affairs, Consumer Cooperatives, and Bismuth 5. Description of the invention (17) 1 I Expensive R only uses its functional chip in a few occasions. Although the overload area 1 1 0 of the invention I is bit K Outside the chip, the access area of the overload area 11 〇 is M type 1 1. An efficient way to send a number of bytes of information at a time. Read first read 1 This is compared to the transmission Column structure 9 writes or captures __ only once in the queue. The 1 input data is not the same. 0 '' Cause of attention 1 1 During the operation 1 If there is a large number of data reached the output queue 74 t 1 1 These input data are put in the overload area U 0 Μ to avoid the overload item on the chip 7 8 and then 4 f to install I 0 Therefore, using the queue structure of the present invention, most of the actions of dropping the frame can be avoided. 〇 At the same time, the capacity of the overload area 110 can be easily adjusted by changing the size of the external memory 36 > and 4 the overload area 1 1 of a special size 1 1 can be assigned and customized. The size of the queue, without the need to affect the performance of the queue 1 1 out of the queue 74 订 order '1 typically * The queue is-~ 1 structure with a sequence • has the structure of the first 1 1 out 0 Typed queues such as »Recycling queue 98 and empty reading 1 j 媛 槽 104. Input The order of the data is not important. If there is 跬 1, the data can be sent directly from the input terminal 1 0 0 to the capture terminal 1 2. The present invention allows information to be transmitted from 1 I 瑄 Article Road 9 卽 Bypass the overload area for queue 〇P want this information Not 1 1 f is order-sensitive (0 rder -S e η t ί ν e), even if the information is in the relevant 1 1 overload zone »it is allowed to do so. Ο For example, the m punch of the recovery zone is not I order-sensitive t can work this way, because when the buffer no longer needs to store 1 '1, the frame will eventually return to the free JStr slot-. 丨. \ Read the table (free 1 i S t * 1 1), so any order in the punch is acceptable. Therefore, when the data is not order-sensitive > If there is a recovery queue 98 written to the external memory 1 I system. Πί Overload area 1 1 0, in order to avoid changes in bandwidth, R should be taken to the end 1 1 17 This paper size applies the Chinese National Standard (CNS) Λ4 specification (210X 297 mm) 9 1472 453 A7 B7 ^ Ministry of History Standards ^ 扃 Consumer Cooperative Cooperative Printing Co., Ltd. 5. Description of the invention (1 8) 1 I 1 0 2 If there is a vacancy, the information will be sent directly from the writing end 100 to the reading end 102 1 | 〇 Recovery ΐ 9 9 8 is __. 伫 伫 f * > 1 Example of the queue of this kind is not sensitive to order 1 The η material <-> in many applications ♦ There are still many order-insensitive 1 1 data types i so> The color holding of the present invention is to find a sort of queue between applications to read back 1 1 t permutation This is a different data type 1¾ I of 1 Note I The example of the actual cell of the present invention is described in Figures 1 and 2 of the multi-port exchange device. Set 1 Set a total of 28 output queues (each — * 涸 is related to — output ports), then fill in 1 of 24. One for 10Hb / s user port »2 涸 For 100 user port for writing this page | Use t One for management port» — One for expansion calendar port (? Dangxian, · 1 I box indicator Lined up for transmission * Output queues 74, 7 5 77 are provided to 1 1 Frame indicators are a place for temporary storage 0 This is the first-in-first-out (FIFO) 1 1 The ordering method is extracted first through FIFO 70 according to the port Indicated in, write frame index 1 to different output queues 74 7 5 > 77 of — this form of port vector 1 1 I 〇 different output queues in the preferred embodiment of the present invention 74 75 N 77 1 1 Includes many or all of the following fields: Single copy bit frame indicator 1 I control opcode or control signal 9 and VLAN (virtual area network) tag, 1 1 I single copy bit mode of the frame »so that the frame will only be sent to a single tilt output I 1 port 0 frame indicator Go to the frame on the external memory 馎 36. Control the operation code identification 1 1 Don't special information related to the frame (newly learned frame) 0 The control signal is controlled by the information from the j 1 operation code 1 to indicate How does this iih port 1 master the frame before transmitting the frame? The VLAN tag is provided to the output frame that must be attached (if the cadmium requires 1 I> VLA Ν tag (tag) a reference 1 However »These examples are only a 1 I The present invention can be applied to other different forms of output queues. 1 1 1 8 This paper size is the state standard of China (CNS) A4 (210X297 mm) 91472 〇 c Α7 Β7 .¾¾ Ink of the Men ’s Workers ’Cooperative of the Standard Soap Bureau 5. Invention of Ke Ming (1 9) Field 0 output fj column 7 4 of the first type of internal structure, 10 M b / S output queue is as shown in Figure 5, 10 Mb / s output queue 7 4 holds will be sent to 10 0Mb / The signal of the S port uses round-robin data. In this example, the input m of the queue and the writing end of the queue 7 6 hold 3 2 input data. And the output queue m of this queue takes The terminal 78 holds 16 input data. Each 1 OMb / s output queue 7 4 input data > has a copy bit and a frame indicator. In the example cell example of the multi-port switching device of the present invention, the VLAN tag is not necessary, because there are no VLAN m strips attached to the 10 Mb / s port. The second type of output queue 74 is a " example. The internal structure of the m example, the 100Mb / s_ output queue is shown in Figure 6. The 100Mb / S output queue 74 holds the input data for the frame that will be sent to the 100 Mbps b / s port. The output of these queues The queue writing end 7 6 holds 6 4 input data * and the output queues of these queues 7 8 1 holds 16 input data 0 each fake m input data includes a VLAN tag,- Part of the control opcode, a single copy bit, a frame index r > An example of the external memory 3 6 is shown in Figure 7. The overall capacity of the external memory may be 4Mb, although in other real examples There are different capacities in the external memory 36 as an overload area according to the present invention, allowing as long as simply You can increase or decrease the size of the output queue by changing its external memory 9. The ratio of all the queue structural parts on a single crystal is * M. All the queue capacity is determined by the production of the wafer. 1 Yes Better 0 In order to satisfy the storage m of the exchange device 12, please * A sample of the external memory 36 m is configured with the following areas »Free slot slot overload area 1 20 Read first read the note on the back and then fill in: Loading η 19 This paper size is applicable to China National Standard (CNS) A4 (210X 297 mm) 9 1472 IT-

Vj、Vj,

53q8G 經"部中央標嗥局员工消費合竹社印^ 五、發明説明 ( 2 0 ) 1 I 回 收 佇 列 過 載 區 12 2 . 多複料佇列過載區1 24 t 管 理 埠 輪 J I 出 ft 列 過 載 區 12 6, 供1 ΰ Μ b /' S 及 1 00 Mb/s 的 埠 使 用 的 個 別 1 1 1 輸 出 佇 列 區 128, 撺充_流排埠輸出佇列過載區1 30 r Η IB ϋ 1 先 計 m 器 )3 2, 及全區訊框嫒衝槽( g 1 〇 b a 1 f r a me b u f f e r I 讀 1 背 1 Ρ 〇 ο 1 )1 3 4 〇 面 I 之 1 注 丨 在 晶 Η 上 > 暫 存 器 6 0 群中 的 記 憶體 基 底 位 址 (b as e 意 | 事 1 I a ό dr e s S ) 暫 存 器 其 全 部 記憶 體 區 的基 底 位 址 (b a s e Jn 再 ά 1 a d d r e s S ) 是 可 規 創 的 〇 在 外部 記 憶 體對 應 U e m or y ϊη h ρ) 中 本 装 頁 1 的 每 __. 塊 區 域 的 基 底 位 址 是可 規 削 的, 且 不 須 要 長 度 暫 存 、- 1 I 器 (1 e η g t hr e g is t e Γ ) 對 應任 何 一 個區 域 長 度 等 於 由 一 基 1 1 I 底 位 址 到 下 個 基 底 位 址 的長 度 0 1 訂 因 各 別 過 載 區 的 長 度 (及容量) 是可 規 割 的 * 所 Μ 每 —* 1 佇 列 的 整 體 容 最 是 可 規 副 的。 本 發 明的 這 些 特 性 允 許 訂 製 1 1 一 交 換 裝 置 當 需 要 時 4 使其 提 供 較大 容 量 的 特 殊 輸 出 佇 1 1 列 〇 1 冰 下 列 過 載 區 儲 存 的 輸 入資 料 並 不適 於 晶 片 12 的 控 制 佇 1 | 列 1 因 而 放 入 外 部 記 憶 體 36 〇 空 閒 媛衝 過 載 區 1 2 0儲存位 1 1 I 址 指 標 到 § 前 未 用 的 全 域 訊框 訊 框 鍰衝 槽 13 4 , t@ I叱 ί佇列 1 1 載 區 12 2儲存訊框指標到不需要的辣结串列(ii n k e d 1 1 1 i St C h a in s ) P 多 複 製 佇 列過 載 區 1 2 4儲存訊框指標, 伴 丨 隨 著 複 製 碼 "> =1 ” (表示排列訊框指標) t 或 伴 随 著 複 製 碼 ,| ” 1 " ( 表 示 成 功 地 傳 輸 訊 框)。 1 r 列 過 載 區 儲 存 不 適 於放 人 晶 片12 t 供 輸 出 m 衝 器 使 1 I 用 的 輸 入 資 料 , η 管 理 埠 輸出 佇 列 過載 區 1 2 6儲存等待傳 1 1 本紙張尺度適用中國國家榡準(CNS ) A4規格(210X297公釐) 20 9 1472 Λ7 B7 五、發明説明(21 ) 輪到管理埠的訊框指標。輸出佇列過載區1 2 8儲存等待要 傳送到適當的1 0 M b / s及i 0 0 H b / s的訊框指標掮充歴流排 埠輸出佇列過載區1 3 0儲存等待要傳送到擴充陳流排埠的 訊框指標。 Μ 1 B計數器區1 3 2包含所有個別埠的統計量,該統計最 會由交換裝置12作周期性地更新。交換裝置12維持8位元 及1 6位元計數器在晶片上,用來儲存Η I Β的统計量。交換 裝置1 2使用一避免Μ I Β資料道失的頻率,Μ更新在外部記 憶體36上之32位元或64位元ΜΙΒ計數器。 全域訊框媛衝槽134包含辣結表列中的鍰衡器,該辣 結表列儲存著已接收的訊框資料。在任何時間,這些辣结 表列包含正確的訊框資料,作_的鍰衡器將由鍰衝器管理 員7 2歸還到空閒緩衝槽1 0 4 ,或者歸鵰到P C I主槠處理器2 8 〇 現在參考到第8_,本發明的一個範例實胞例,由任 何MAC埠或由PCI匯流排接收的訊框資料,Μ辣结表列的型 -I - - - ^—^1 ' - - - I— I- J^i 士皮._ 1--- - i i ^—^1 I I I (讀先閱讀背面之注意事項再"朽本頁) 有 例 施 實 冏 不 明 發 本 然 雖 ο 中 6 3 S 憶 記 部 外 在 存 儲 態 Μ滴部中央標準局負工消費合作社印絮 衝 0 的 列 表 结 辣 成 造 來 用 被 中 例 本 但 度 長 器 β 8 媛 的 同 不 被 標 指 元 位 的 ο 4 1± 器 Β 緩 此一一 這 組 元 位 6 5 2 為 度 畏 其 ο 4 器 中 B 8 鍰 框 訊 到 收 接 埠 10個槽一 器何 衝 任 鍰的 閒中 r . 2 空 1 的 置 12裝 置 換 裝 交 換在 交0 於 一 存 儲 結 鏈 來 用 標 指 址 位 來 送 4 ο 器 樣 鍰 閒 空 求 要 2 7 員 理 管 器 的 器 衝 緩 1 第 的 6 3 槽 憶 記 部 LT 夕 0 框 訊 該 存 儲 ο 4 11 標 指 址 位 標 指 框 訊 的 框 訊 sg 成 變 在 現 框 訊 著 存 儲 來 原 標 位 本紙張尺度適用中國國家標準(CNS ) Λ4規格(210Χ 297公釐) 2 1 9 1472 4 53 053q8G Ministry of Standards and Bureau of the People ’s Republic of China ’s Consumers ’Consumption of the Hezhu Society ^ V. Description of the Invention (2 0) 1 I Recycling queue overload zone 12 2. Multiple compound queue overload zone 1 24 t Management port wheel JI out ft Overload zone 12 6, Individual 1 for 1 μM b / 'S and 100 Mb / s ports 1 1 1 Output queue zone 128, charge_flow port output queue overload zone 1 30 r Η IB ϋ 1 first counts m) 3 2 and the whole frame frame gutter (g 1 〇ba 1 fra me buffer I read 1 back 1 Ρ 〇ο 1) 1 3 4 〇 face I 1 Note 丨 in the crystal上> Register 6 0 base address of the memory in the group (b as e intention | 事 1 I a ό dr es S) register the base address of all its memory areas (base Jn re-sub 1 addres S) is configurable. It is installed in the external memory corresponding to U em or y ϊη h ρ). The base address of each __. Block area on page 1 is configurable and does not require temporary storage.-1 I device (1 e η gt hr eg is te Γ) corresponds to the length of any area equal to a base 1 1 I The length from the bottom address to the next base address 0 1 The length (and capacity) of each overload area is configurable * Μ Each — * 1 The overall capacity of the queue is the most regulable of. These features of the present invention allow the custom 1 1-exchange device to provide special output with larger capacity when needed 4 1 column 0 1 input data stored in the overload area below the ice is not suitable for the control of the chip 12 1 | Column 1 is thus placed in external memory 36 〇 Idle element rushes into the overload area 1 2 0 Stores the bit 1 1 I address pointer to the § previously unused global frame frame 锾 slot 13 4, t @ I 叱 ί 伫 Column 1 1 Load area 12 2 Store frame indicators to unwanted spicy knots (ii nked 1 1 1 i St C ha in s) P Multi-copy queue overload area 1 2 4 Store frame indicators, with Code " > = 1 "(indicating frame index) t or accompanied by copying code, |" 1 " (indicating successful transmission of frame). The 1 r row overload area stores the input data that is not suitable for putting the chip 12 t for the output m. The input data used by the punch 1 1, η management port output queue overload area 1 2 6 storage waiting for transmission 1 1 This paper size is applicable to Chinese national standards (CNS) A4 specification (210X297 mm) 20 9 1472 Λ7 B7 V. Description of the invention (21) Frame indicator of the turn management port. Output queue overload area 1 2 8 Storage waiting to be transmitted to the appropriate 1 0 M b / s and i 0 0 H b / s frame indicator. Filling the flow port output queue overload area 1 3 0 Storage waiting to be Frame indicator sent to the extended Chenliu port. The M 1 B counter area 1 2 2 contains statistics for all individual ports, and the statistics are most periodically updated by the switching device 12. The switching device 12 maintains 8-bit and 16-bit counters on the chip and is used to store the statistics of Η I Β. The switching device 12 uses a frequency that avoids data loss of the M I B data, and M updates the 32-bit or 64-bit M I B counter on the external memory 36. The global frame element slot 134 contains the weighing instrument in the spicy knot list, which stores the received frame data. At any time, these tables contain the correct frame information. The balancer will be returned by the administrator 7 2 to the free buffer slot 104, or the PCI host processor 2 8. Reference is now made to Section 8_, which is an example cell example of the present invention. The frame data received by any MAC port or by the PCI bus is listed in the following table: -I---^ — ^ 1 '--- I— I- J ^ i 士皮 ._ 1 ----ii ^ — ^ 1 III (Read the precautions on the back first and then quoting this page) There is an example of implementation, but it is unknown if it was sent ο Middle 6 The external storage state of the 3 S memory department, the M drop department, and the Central Standards Bureau ’s Consumer Cooperative Cooperative Printing Co., Ltd. printed a list of 0. It was created to be used in the example, but the length of the device β 8 yuan is not labeled as a bit. Ο 4 1 ± device B slow down one by one this group of bits 6 5 2 is the degree of fear ο 4 device B 8 锾 frame message to the receiving port 10 slot one device He Chong Ren 锾 leisure r. 2 empty 1 sets of 12 devices are exchanged at the intersection of 0 and a storage link to use the index address to send 4 ο sample 2 7 members of the organ manager to slow down 1st 6 3 slot memory LT Xi 0 frame message storage ο 4 11 index address frame pointer frame message sg changed into the current frame message storage The original paper size is in accordance with the Chinese National Standard (CNS) Λ4 specification (210 × 297 mm) 2 1 9 1472 4 53 0

G A7 B7 經济部中央標牟局KK工消费合作社印^ 五、發明説明(2 2 ) 訊框指標係使用於交換裝貿次系統56中,用來排列將要傳 輸的訊框〇 壊浙器1 4 0 W媛衝器表頭1 4 2的位址指標辣結茌一起. 緩衝器表韻1 4 2會指向下一個在記憶體中的緩衝器。嫒衝 器頭部Μ 2同時含有在緩衝器14 0中的訊框資料的資訊。第 一個媛衝器表頭有1 2個位元組,如第9 a圖範例緩衝器的格 式所示。下一個緵衝器表頭是4位元組,如第9h圖所示。 外部記憶體組發區有2個記憶體槽 (batik), 16個位元組的 長度^所以每一個缓衝器訊框的儲存容量為256B-16B = 240 B〇 如第9a闕及第9b圖所示,第一個及第一個K後鍰衝器 的格式,該格式包括以下的各禰: 媛衝器格示位元:指示使用中的緩衝器是何種格示, 如果是”1”的話,表示是第一個锾衝器,該緩衝器是U個 位元組的長度。如果是” 0”的話,表示非第一個媛衝器, 該鍰衝器格式是4位元組。當在辣結這些緩衝器時,這個 規則使用於每一個剩下的鑀衝器。 Ε位元(標記訊框的结束):指示這是一個訊框的最後 鍰衝器,當這個位元被設定為“1” ,該練结串列中就沒 有其它緩衝器。 C位元(C R C錯誤偵測):指示由接收器偵測到一個C R C 錯誤,當C位元被偵測到,傳輸的函數故意傳輸一個倒轉 的C R C ( ί η ν e「t e d C R C U位元(對準錯誤):指示偵測到接收 訊框的訊框對準錯誤(伴隨著一個C R C錯誤)。 (請先_閲讀背而之注意事項再填巧本頁) 裝 •I 水 本紙張尺度適用中國國家標準(CNS ) Λ4规格(2丨0X 297公釐) 22 91472G A7 B7 Printed by the Central Bureau of Standards, Ministry of Economic Affairs, KK Industrial and Consumer Cooperatives ^ V. Description of the Invention (2 2) The frame indicator is used in the exchange equipment sub-system 56 to arrange the frames to be transmitted. The address index of the 4 0 Wong punch head 1 4 2 is hot together. The buffer table rhyme 1 4 2 will point to the next buffer in the memory. The buffer head M 2 also contains information of the frame data in the buffer 140. The first header has 12 bytes, as shown in the format of the example buffer in Figure 9a. The next punch header is 4-byte, as shown in Figure 9h. The external memory block has 2 memory slots (batik), with a length of 16 bytes ^ so the storage capacity of each buffer frame is 256B-16B = 240 B. Such as 9a 阙 and 9b As shown in the figure, the format of the first and first K posterior punches, the format includes the following elements: Yuan punch format: indicates the format of the buffer in use, if it is " "1" means that it is the first buffer, and the buffer is the length of U bytes. If it is "0", it means that it is not the first elementary punch, and the format of the bit punch is 4 bytes. When freezing these buffers, this rule applies to every remaining punch. E bit (marks the end of the frame): indicates that this is the last punch of a frame. When this bit is set to "1", there are no other buffers in the training string. C bit (CRC error detection): indicates that a CRC error was detected by the receiver. When the C bit is detected, the transfer function intentionally transmits an inverted CRC (ί η ν e "ted CRCU bit ( Misalignment): indicates that a frame misalignment of the received frame was detected (with a CRC error). (Please read the precautions on the back before filling out this page.) Loading • I Water-based paper size applies China National Standard (CNS) Λ4 specification (2 丨 0X 297 mm) 22 91472

®濟部中央標準局员工消贽合竹社印采 五、發明説明(2 3) 0位元(接收FIFO過載):指示接收FIFO已過載,在猨 衝器中的資料是不正確的。 緩衝器長度:媛衝器表頭之後的第一個位元組開始算 起,鍰衝器資料锎位中正確的位元組數目。此畏度不可包 括移位位元組值。 下一媛衝器指標:該指摞指到下一個媛衝器^當E位 元設定為1之後,下一緩衝器指標則非正確值。 移位位元組(〇 f f s e t b y t e )計數:指示媛衝器訊框資 料區,訊框的第一個位元由何處開始,零移位量表示該資 料將會在緩衝器表頭1 4 2之後的第一個位元開始,零移位 最表示該資料將會在緩衝器的第1 6個位元開始。一個非零 的移位位元,訊框資料將由媛衝器的最開顗算起16B +移位 罱之後開始p傳輸函戲(transmit function)將會跳過移 位位元桶所記載的位元組數目。 P位元(埠型態):指示新進人的接收訊框的埠型態。 零表示10Mb/S,賣的話表示是100Hb/s。此位元是規劃交 換裝置1 2在訊框完全接收之前,轉送訊框到擴充匯流排, 並將外部記憶髖36當作緩衝器使用,將訊框送到外部記憶 體36時,此位元由主櫬28聯合時間戳記一起運作。 T位元:指示已接收訊框的型態:搮籤(tagged)或未 加標籤Untagged)。壹表示標籤訊框,及VUN識別櫊包含 已接收VLAN ID。零表示未加標籤訊框,及VLAN ID是錯誤 的〇 接收埠號碼:埠的號碼在該埠訊框將被接收 --------裝-----"—訂------\ (請先閲讀背而之注意事項再填寫本頁) 本纸張尺度適用中國國家榇率(CNS ) A4規格(210X 297公釐) 23 9 1472 0 8 Λ7 B7 五、發明説明(24 ) V L Λ N識別碼:由加標镝埠接收的V L A N I D 。如果訊框 ώ 一非加標糖増(u n t a g g e d ρ ο r ί )接收的話,Μ這個機位 是錯誤的。 R位元(再計算C R C ):指示在傳輸函數中的C R C必需拆 散及重算,當一個加標籤訊框已經接收後,交換裝置1 2設 定該位元為1。再者,如果主機28修改該訊框的内容,主 機28必須設定這個位元為1。當交換裝置12傅輸一個訊框 時,它將會檢査這個位元Μ決定是否要傳送現存的CRC或 者拆散並重算該C R C。 Α位元(附加CRC):表示訊框資料的尾部沒有CRC,主 機28可以在記憶體中產生一個訊框(此訊框沒有CRC),然 後設定此一位元為1。當傳輸該訊框時,交換裝置12將會 產生及附加一個CRC。如果A位元已被設定,則訊框不可含 有 CRC。 F位元(格式位元):辨別訊框長度/時間戳記播位。零 表示是該權位是新進人訊框的時間戳記,壹的話表示該襴 是位已接收訊框的訊框長度。 訊框長度/時間戳記:決定在F位元。如果F位元清除 ,該欄位代表來自已接收訊框開頭的時間戳記。時間戳記 有lms的解析度。如果F位元被設定為1,此檷位表示已接 收訊框的總長度,此長度包栝CRCK及任何已接收的 VLAN 標籤。當一訊框被接收後,交換裝置1 2使用時間戳記(來 自計時器暫存器)標記此檷位。如果主機28已規副交換裝 置12在訊框完全接收前,轉送訊框至擴充匯流排,它可使 24 ----- -----_ ___τ (讀先Μ讀背而之注意事項再填寫本頁) 經Μ部中央標準局員工消費合仃.杜印挈 本紙張尺度適用中國國家#率(CNS } Λ4規格(2!0><297公釐) 9 1472 4 5 經漪部中央標準局員工消f合作杜印製 五、發明説H/ ( 2 5 ) 1 t 用 時 間 戳 來 最 測 (伴陳著該接收埠的速度) 它 可 由 外 部 記 J I 憶 體 3 G 抓 取 多 少 資 料 1 而 不 用 過 度 讓 取 訊 框 —A·· 貝 料 〇 一 旦 所 1 1 1 | 有 訊 框 被 接 收 後 » 交 換 裝 置 12將 訊 框 的 長 度 寫 入 此 m 位 I 諳 I * 並 設 定 F 位 元 為 1 〇 凡· 聞 I 讀 1 | 複 m 號 碼 用 來 表 示 由 埠 向 量 F I F 0 7 0 排 列 成 功 » Μ 供 ή 而 之 1 1 傳 輸 的 複 製 數 i 如 果 緩 衝 器 管 理 員 72 需 要 在 多 複 製 快 取 注 意 1 1 事 1 96 中 留 位 置 給 新 的 輸 人 資 料 t 此 m 位 用 來 儲 存 訊 框 指 標 的 項 再 1 填 複 製 數 巨 0 本 百 裝 | 第 1 0 圖 詳 细 描 述 第 3圖交換裝置次系統56某些元件。 1 1 I 這 元 件 提 供 媛 衝 器 儲 存 空 間 > Μ 便 用 來 儲 存 訊 框 ♦ 及 回 1 1 I 收 這 m iff 器 « 當 旦 瑄 m 衝 器 不 再 儲 存 這 些 訊 框 時 t 1 訂 1 再 便 用 這 1¾ 緩 衝 器 〇 如 前 所 述 t 所 有 的 輸 出 佇 列 7 4 9 75 傳 遞 訊 框 指 標 到 m 衝 器 管 理 員 (除了輸出佇列7 7M外) * 媛 1 1 衝 器 管 理 員 72 會 為 這 些 被 訊 框 指 標 指 到 的 訊 榴 排 時 程 » 媛 1 | 衝 器 管 理 員 控 制 下 列 功 能 : 1) 管 理 交 換 裝 置 12的 內 部 匯 流 1 排 Ο 2 ) 便 利 排 序 訊 框 指 標 到 輸 出 佇 列 74 1 及 由 輸 出 佇 列 74 1 | 反 排 序 訊 框 指 m 〇 3) 管 理 控 制 佇 列 90 > 98 » 用 以 配 置 及 歸 1 ! 堪 鍰 衝 器 到 空 閒 媛 衝 器 10 4〇 4 )控制到外部記憶體3 6的資 1 1 料 流 > 及 從 外 部 記 憶 體 3 6 來 的 資 料 流 〇 5 ) 維 持 記 憶 體 的 結 1 1 構 包 括 Ml Bs 及 過 載 區 〇 m jSR* 衝 器 管 理 員 72包 括 一 個 排 程 的 .1 1 功 能 » 以 配 置 所 有 外 部 記 憶 體 36的 存 取 9 這 些 存 取 包 括 1) 1 寫 一 已 接 收 訊 框 資 料 到 憶 體 緩 衝 器 1 40 〇 2 )讀取來自記 1 1 I 憶 體 m 衝 器 1 4 0的訊框資料, 以供傳輸。 3 )維持(即 讀 寫 ) 1 1 I 輸 出 佇 列 7 4 及 控 制 佇 列 9 0 98 的 每 ~~1 傾 過 載 區 中 的 訊 框 指 1 1 本紙張尺度適用中IS國家標率(CNS)Λ4規格(2〗0X297公釐) 25 9 1472®Employees of the Central Standards Bureau of the Ministry of Economic Affairs of the People's Republic of China. 5.Invention description (23) 0 bit (receive FIFO overload): indicates that the receive FIFO is overloaded, and the data in the buffer is incorrect. Buffer length: Starting from the first byte after the header of the puncher, the correct number of bytes in the puncher data bits. This degree of fear cannot include shifted byte values. Next element indicator: This index means the next element indicator. When the E bit is set to 1, the next element index is not correct. Shift byte (〇ffsetbyte) count: indicates the data area of the frame of the yuan punch, where the first bit of the frame starts, a zero shift indicates that the data will be in the buffer head 1 4 2 The first bit after that starts. Zero shifting means that the data will start at the 16th bit of the buffer. A non-zero shift bit. The frame data will be calculated from the maximum opening of the yuan puncher. After 16B + shifting, the p transfer function will skip the bits recorded in the shift bit bucket. The number of tuples. P bit (port type): indicates the port type of the incoming frame of the newcomer. Zero means 10Mb / S, and when sold, it means 100Hb / s. This bit is planned for the switching device 12 before the frame is completely received, and it transfers the frame to the expansion bus and uses the external memory hip 36 as a buffer. When the frame is sent to the external memory 36, this bit It works together with the main time stamp 28. T bit: indicates the type of the received frame: tagged or untagged. One indicates the tag frame, and the VUN identification contains the received VLAN ID. Zero indicates that the frame is not tagged and the VLAN ID is wrong. 0 Receive port number: The port number will be received in the port frame. -------- Install ----- " — 订- ---- \ (Please read the precautions before filling this page) This paper size is applicable to China National Standard (CNS) A4 specification (210X 297 mm) 23 9 1472 0 8 Λ7 B7 V. Description of the invention (24) VL Λ N identification code: the VLANID received by the standard port. If the message frame is received by a non-standardized sugar puppet (u n t a g g e d ρ ο r ί), the M position is wrong. R bit (recalculate C R C): indicates that C R C in the transfer function must be disassembled and recalculated. When a tagged frame has been received, the switching device 12 sets the bit to 1. Furthermore, if the host 28 modifies the content of the frame, the host 28 must set this bit to one. When the switching device 12 enters a frame, it will check this bit M to decide whether to transmit the existing CRC or disassemble and recalculate the CR C. Bit A (additional CRC): It means that there is no CRC at the end of the frame data. The host 28 can generate a frame in the memory (the frame does not have a CRC), and then set this bit to 1. When transmitting the frame, the switching device 12 will generate and append a CRC. If bit A is set, the frame must not contain a CRC. F bit (format bit): identify the frame length / time stamp broadcast bit. Zero indicates that the weight is the timestamp of the newly entered frame, and one indicates that the frame is the frame length of the received frame. Frame length / time stamp: Determined in F bits. If the F bit is cleared, this field represents the timestamp from the beginning of the received frame. Timestamp has lms resolution. If the F bit is set to 1, this bit indicates the total length of the received frame. This length includes the CRCK and any received VLAN tags. When a frame is received, the switching device 12 uses a time stamp (from the timer register) to mark this bit. If the host computer 28 has the secondary switching device 12 transmitting the frame to the expansion bus before the frame is completely received, it can make 24 ----- -----_ ___ τ Please fill in this page again) Consumption of employees of the Central Bureau of Standards of the Ministry of Education. Du Yin's paper size is applicable to China's national #rate (CNS) Λ4 specifications (2! 0 > < 297 mm) 9 1472 4 5 Jingyi Department The staff of the Central Bureau of Standards cooperated with Du Du. V. Invented that H / (2 5) 1 t was measured with a time stamp (with the speed of the receiving port). It can be captured by an external JI memory 3G. Data 1 without excessively letting the frame—A ·· beautiful. Once all 1 1 1 | After a frame is received »Switching device 12 writes the length of the frame into this m bits I 谙 I * and sets F The bit is 1 〇 Fan · wen I read 1 | The complex m number is used to indicate that the port vector FIF 0 7 0 is successfully ranked »Μ price and 1 1 is the number of copies i transmitted if The buffer manager 72 needs to pay attention to the multi-copy cache 1 1 event 1 96 to reserve a place for new input data t This m bit is used to store the frame index and then 1 fill the number of copies 0 0 hundred | 1 0 The detailed description of some components of the exchange system sub-system 56 in Figure 3. 1 1 I This component provides storage space for the puncher > Μ is used to store the frame ♦ and return 1 1 I to receive the m iff device « When the 瑄 m punch no longer stores these frames, order 1 and then use the 1¾ buffer. As described above, all output queues 7 4 9 75 pass the frame indicators to the m punch manager ( In addition to the output queue 7 7M) * Yuan 1 1 The punch manager 72 will schedule the grenade schedules indicated by the frame indicators »Yuan 1 | The punch manager controls the following functions: 1) Management of the switching device 12 Internal confluence 1 Ο 2) Conveniently sort frame indicators to output queue 74 1 and output queue 74 1 | Reverse frame indicator refers to m 〇3) Management control queue 90 > 98 »Used to configure and return to 1! Device to idle device 10 4 0 4) control the data flow to external memory 3 6 1 1 stream > and the data flow from external memory 3 6 0 5) maintain the memory structure 1 1 structure including Ml Bs and overload area 0m jSR * The Blast Manager 72 includes a schedule. 1 1 Function »To configure access to all external memory 36 9 These accesses include 1) 1 Write a received frame data to the memory The body buffer 1 40 〇 2) read the frame data from the memory 1 1 memory m 1 14 0 for transmission. 3) Maintenance (ie, reading and writing) 1 1 I output queue 7 4 and control queue 9 0 98 each ~~ 1 frame in the overload zone refers to 1 1 IS national standard (CNS) in this paper Λ4 specifications (2〗 0X297 mm) 25 9 1472

經浐部中央標尊局負T_消費合作杜印裝 五、發明説明( 2 6 ) 1 1 標 〇 4 )更 新 Μ 1 B計數器。 1 緩 衝 器 管 理 員 把 一 指 定 訊 框 複 製 到 所 有 適 當 的 輸 出 庁 1 1 列 7 4 75後 ♦ 埠 向 最 K I P 0 7 0 si- 算 複 製 的 數 目 (複製號碼) * 1 I Μ 及 放 置 訊 榧 指 標 和 複 製 號 碼 到 多 複 製 佇 列 9 0的 書 寫 端 92 先 閱 讀 背 面 之 +丨 I η 複 製 號 碼 若 是 零 * 表 示 該 訊 框 是 不 可 轉 送 的 若 是 壹 的 I 1 話 1 表 示 是 — 個 單 複 製 傳 輸 若 是 該 號 碼 大 於 費 t 則 表 示 意 事 1 1 是 —1 個 多 複 製 傳 輸 t 此 三 個 例 子 將 在 下 面 解 說 0 .當複 再 填 1 ά I 製 碼 是 零 時 表 示 訊 框 指 標 有 一 個 零 轉 送 埠 向 量 9 且 沒 巧 本 ii 有 任 何 位 元 被 設 定 f 埠 向 量 F夏F 0 7 0直 接 傳 遞 訊 框 指 檷 到 在 1 1 I 回 收 佇 列 9 8的 書 寫 端 10 〇 〇 當埠向量FIF070服務回收佇列 I 1 1 98時 » 這 種 情 形 將 敘 述 於 下 0 鍰 衝 器 管 理 員 72打 斷 媛 衝 器 r 1 的 鰱 结 串 列 > 歸 堪 每 一 個 工 閒 (f re e)緩 衝 器 的 位 址 指 標 到 訂 1 空 閒 m 衝 榷 104的書寫端1 06 〇 1 1 當 複 製 號 碼 是 看 時 * 表 示 是 一 個 單 複 製 傳 輸 » 埠 向 1 1 最 F I F 0 7 0 複 製 訊 框 指 標 » 控 制 信 號 /控制操作碼, 及 VLAH 1 ,冰 指 標 到 適 當 诲 的 輸 出 佇 列 7 4 埠 向 量 F IF 07 0 設 定 輸 出 佇 列 I 1 7 4中 的 輩 複 製 位 元 (見第5圖 及 第 6圖) K 表 示 這 是 一 個 單 1 1 I 複 製 傳 輸 當 緩 衝 器 管 理 員 72由 該 埠 輸 出 佇 列 74 _ 取 訊 框 1 i 指 標 及 輩 複 製 位 元 時 * m 衝 器 管 理 員 72會 如 前 所 述 的 做 傳 I 1 輸 的 排 程 〇 緩 衝 器 管 理 員 72使 用 訊 框 指 標 配 置 在 外 部 記 憶 :丨 1 體 3 6中 的 第 一 個 緩 衝 器 r 此 媛 衝 器 儲 存 著 訊 框 〇 緩 衝 器 管 理 員 72請 取 第 —* 個 緩 衝 器 由 第 一 個 媛 衝 器 獲 取 資 料 , 並 I 送 該 資 料 到 適 當 的 MAC傳輸 F IF0 54 〇 假 設 該 訊 框 擴 及 多 個 1 i 緩 衝 器 * 上 面 所 述 的 m 结 會 m 结 到 下 一 個 緩 衝 器 f 如 此 的 1 1 本紙張尺度適用中國國家標準(CNS ) Λ4規格(210Χ 297公釐) 26 9 1472 ό ο B7 經濟部中央標準局負工消费合作社印繁 五、發明説明 ( 2 7 ) 結 構 提 供 m 衝 器 管 理 員 7 2 — 個 位 址 * 可 用 Μ 尋 找 及 傳輸跟 該 訊 框 有 關 鍵 結 的 所 有 緩 衝 器 0 一 旦 該 資 料 被 放 在 FIFO 54Μ 供 傳 輸 後 則 該 緩 衝 器 將 作 廢 而 歸 遨 給 空 閒 m 衝槽 10 4 , Μ梗重新配置而提供其他的訊框儲存資料。 當 複 η 號 m 大 於 青 時 * 埠 向 最 K I F0 7 0 複 製 訊 框 指標, 控 制 信 號 /控制操作碼, 及V L A N指標到適當埠的輸出佇列 7 4 〇 (這是提到佇列7 4的情況, 如果提到佇列Ί 5 , 77也是 同 樣 的 情 形 )0 埠向最f I FQ7 0清除輸出佇列74中適當的訊 框 指 標 中 的 單 複 製 位 元 t 以 複 製 號 碼 大 於 壹 的 狀 態 ,將訊 框 指 標 放 人 多 複 製 佇 列 90 的 書 寫 端 9 2 〇 Ρ 要 緩 衝 器 管 理 員 72 讀 取 個 訊 框 指 標 及 茌 輸 出佇列 74其 中 之 一 清 除 單 — 複 製 位 元 時 f 媛 衝 器 管 理 員 會 排程訊 框 的 傳 輸 t 同 時 檢 査 多 複 製 快 取 96 * 看 輸 入 資 料 的 訊框指 標 是 否 有 複 製 號 碼 為 1者, 如果在多複製快取96中發現有 複 製 號 碼 為 1的訊框指標。 媛衝器管理員72跟訊框的單複 m 傳 輸 —1 樣 * 排 程 供 傳 輪 用 的 訊 框 * 且 在 傳 輸 過 程 中回收 該 m 衝 器 ίΜ 是 f 如 果 訊 框 指 標 沒 有 在 多 複 m 快 取 9 6中, 或 者 在 多 複 製 快 取 96 中 的 訊 框 指 標 大 於 1 , 媛衝器管理員 72 會 傳 輸 該 訊 框 但 並 不 回 收 緩 衝 器 〇 傳 輸 成 功 之 後 ,媛衝 器 管 理 員 72放 置 該 訊 框 指 標 的 一 份 副 本 > 隨 著 複 製 號碼“ -1 ? 到 多 複 製 佇 列 90 輸 出 端 92 0 當 每 次 多 複 製 訊 框 傳 輪 的 時 候 * 只 要 緩 衝 器 管 理員72 沒 有 在 多 複 製 快 取 96 中 發 規 有 複 製 號 碼 為 1的訊框指標, m 衝 器 管 m 員 7 2 會 放 置 份 訊 框 指 標 的 複 製 品 到 多 複製佇 本祕尺度通用中國國家標準(峰(別x贈楚)2? 91472 °S〇 B7 經消部中央標::f-局負工消f合作社印繁 五、發明説明 ( 2 8) 1 1 列 9C 中 因 此 t 任 何 時 間 * 多 複 製 貯 列 90將 包 含 著 有 複 制 Ί 號 碼 .) 1 ί 内訊框指標 或是需多複製號碼- 1的訊框指標副 ί 本 〇 二 讀 ! 1 m 衝 器 管 理 員 7 2 經 常 服 務 多 複 製 佇 列 90 1 及 多 複 製 快 閱 讀 I 背 1 取 96 K 便 m 收 作 m 的 m 衝 器 當 它 腋 務 多 複 製 ί宁 列 90 f 讀 面 1 注 1 取 有 複 m 碼 *1 > 1 ” 的 訊 抿 指 m 锾 衡 器 管 理 員 霄 試 將 此 意 事 1 項 | 新 輪 人 資 料 (訊框指標和複製號碼) 放 入 多 複 製 快 取 96 〇 如 再 J/i 果 多 複 製 快 取 96滿 * 緩 衝 器 管 理 員 72會 留 空 間 給 新 的 訊 ή 本 頁 裝 1 框 指 標 〇 緩 衝 器 管 理 員 72 謓 取 一 較 舊 的 輸 入 資 料 » 更 新 在 1 1 外 部 記 憶 體 36 中 媛 衝 器 表 頭 中 此 輸 人 資 料 的 複 製 號 碼 » 1 \ 然 後 由 多 複 製 佇 列 96 清 除 此 較 舊 的 輸 入 資 料 〇 如 果 在 多 複 [ 1 訂 製 快 取 96 中 尚 有 空 間 可 利 用 擎 緩 衝 器 管 理 員 7 2可 將 新 的 輸 1 1 入 資 料 由 多 複 製 佇 列 90 1 送 至 多 複 製 快 取 96 中 0 ! ! 緩 衝 器 管 理 員 72服 務 多 複 製 佇 列 90 * 並 讀 取 複 製 號 碼 1 1 -1 的 訊 框 指 標 t 它 會 搜 尋 多 複 製 快 取 96 4 尋 找 複 製 號 碼 ί ,VJC 1” 的訊框指標位址。 Μ便於減少或刪除。 如果緩衝器 1 1 管 理 員 7 2 發 m 一 個 U > = 的訊椐指標, 緩衝器管理# 72 1 I 將 1 ) 如果複製號碼大於1 r 則 減 少 多 複 製 快 取 的 訊 框 指 標 1 1 1 * 或 2 ) 如 果 複 製 Dk& 碼 刚 好 為 1的話, 刪除多複製t) ί取 ‘中 1 1 該 訊 框 指 m /複製數個輸人資料免 i將訊框指標放入回收佇 :丨 1 列 98 〇 1 如 果 m 衝 器 管 理 員 並 未 發 現 複 製 號 碼 t* > =1 ’, 的韶 框 1 1 指 標 鑀 衝 器 管 理 員 72 會 搜 尋 外 部 記 憶 體 36 中 的 訊 框 指 標 ! | m 衝 器 表 頭 (見第9 阃 ), 尋JJ ί « 3符合的複製號硝 0 如果 在 1 本紙張尺度適Λ +國國家標準(CNS ) Λ4現格(2】0X297公釐) 28 91472 3〇已〇 A7 B7 鯉"部中央標準局貝工消贫合竹社印製 五、發明説明( 2 9 ) 1 | 記 憶 體 中 的 複 製 號 碼 是 *4 1 " η 緩衝 器 管 理 員 72會 將 訊 框 指 1 1 標 放 入 回 收 f? 列 98 0 如 果 在 記 憶 體 r|W. 屮 的 號 碼 是 “ > 1 〇 鍰 i i 衝 器 管 m 72放 置 訊 框 指 標 及 複 製 號 碼 到 多 複 m 佇 列 96 * 請 先 1 1 i 然 後 減 少 複 製 碼 〇 閱 讀 1 m 衝 器 管 理 72經 常 Μ 請 取 訊 框 指 檷 的 方 式 來 服 務 回 背 面 1 1 之 ) 注 1 收 佇 列 98 〇 然 後 運 作 (ν a I k i n g )鏈結串列歸堪緩衝器到空 意 f 事 項 再 填 1 I 閒 m 衝 器 槽 10 4 〇 此動作只歸遵有零埠向虽訊框的锾衝器 1 , 且 由 埠 向 最 F I F 0 7 0 排 列 到 回 收 佇 列 0 或 者 歸 遛 有 著 多 複 % 本 頁 裝 1 製 向 最 及 已 全 部 傳 輸 完 成 的 訊 框 0 如 前 所 述 當 單 複 製 訊 i I 框 的 被 傳 輸 之 後 t 單 複 製 訊 訊 用 的 鍰 衝 器 鐽 結 將 直 接 歸 m 1 ί I 到 緩 衝 器 槽 104〇 1 訂 如 果 因 輸 出 佇 列74及 在 外 部 記 憶 體 3 6中 的 a M. m 載 區 11 0 1 已 經 滿 了 致 埠 向 垦 FIFO 7 0不能將有單複製轉送向量 1 1 的 訊 框 指 標 放 入 輸 出 庁 列 74 r 該 訊 框 將 被 拋 棄 0 該 m 框 ! 1 指 標 將 歸 m 到 回 收 佇 列 98 * 該 拋 棄 訊 框 的 ffilr 動 作 被 父 換 裝 置 J 管 理 資 湄 (m an a g e ϊη e η t re SO υ Γ c e 0 f t h e S w 1 t C h)所 紀 錄 Ί I I 〇 如 果 因 為 一 或 多 個 輸 出 佇 列 及 它 們 在 外 部 記 憶 體 36 I 1 1 中 的 過 載 區 1 10已滿. 埠向量F I F 0 70不 能 放 置 —1 或 多 個 具 1 1 多 複 製 轉 送 向 量 的 訊 框 指 標 1 於 是 該 訊 框 只 有 轉 送 到 尚 有 1 1 空 間 的 m 出 佇 列 而 複 m 號 碼 放 入 多 複 製 佇 列 90 • 只 單 獨 1 反 映 出 已 成 功 地 放 置 訊 框 指 標 0 不 能 放 置 (η on -P 1 a c e me n t 1 I )的訊框指標將被供訊框指檩不能排列埠的交換裝置管理 1 1 I 資 源 所 紀 錄 如 果 因 為 . 或 多 個 輸 出 佇 列 Μ 及 它 們 1± 外 部 1 1 本紙張尺度通用中_家標準(CNS ) M卿1〇x™ 2g 7 2 〇8 0 Α7 Β7 經漪部中央標率局負工消贽合作社印^ 五、發明説明(30 ) 記憶體36中的過載區110已滿,埠向最FIP070不能放置一 或多俩具多複製轉送向最的訊框指標,該訊框將傳遞到回 收佇列9 8 , 交換裝置管理資源被正式地通知。 埠向最M F 0 7 0採用可完成上述動作的理輯,例如, 確定哪一俩輸出佇列已滿,只將訊框指摞放入那呰未滿的 佇列而其他Β滿的則放棄,如此的邏輯可由本功能敘述之 技转的基本技巧輕易地提供。輸出庁列滿了與苔,可經由 比較一特別輸出佇列的訊框指標的數目來決定(此數值紀 錄在一個暫存器60中),並與一個可指示此佇列已滿的臨 界值做比較。基於Κ上的認知,對多複製訊框而言,埠向 最F I F 0 7 0只放棄那些將要放進輸出佇列的訊框指標,並 增加已放棄訊框的記數值(保存在別的暫存器60中)。 多複製佇列90是緩衝器管理員72所使用的一個高優先 權佇列,在所有的緩衡器(即位址指標)儲存訊框之前,用 以紀錄一特定的多複製訊框有多少要傳輸要完成,所有的 鍰衝器都可回收到空閒緩衝槽104。此一輸出佇列的書寫 端92及讓取端94個別持有64及16個輸入資料。多複製佇列 90饋送資料到多複製快取96,而多複製快取96被鍰衝器管 理員72用來決定什麽時候該回收鍰衝器,多複製佇列的結 構描述於第12圖。 埠向量FIFO 70,根撺訊框指標成功地放入輸出佇列 7 4的訊框指標數目,複製訊框的訊框指標及“ > 1 ”的複製 號碼到多複製佇列90。如果一皤特定的輸出佇列74已滿, 埠向晕F I卩ϋ 7 0不可放貿訊框指檷的複製品到該輸出佇列 ;紙張尺反適用中國國家栋準(CNS > Α4規格(210X297公釐> 30 91472 (請先W讀背而之注意事項再填寫本頁) ^衣_ 〇8 0 A7 B7 好漪部中央標準局月工消舟合作社印紫 五、發明説明(31 ) 7 4 ;因此它不可以將此一事件當作是成功地確定複製號碼 的例子 。 每次媛衝器管理員72謓取一個輸出佇列訊框指標,並 發現單複製位元為“0"(即此是一個多複製的動作^它 舍尋找多輸出快取中有複製號碼為“1”的訊框指搮,複 製號碼為“ 1 ”的訊框指標表示這是一個最後的傅輸。如 果找到Γ,在傳輸每一個鍰衝器内容之後,媛衝器管理員 7 2經由提供作廢的媛衝器到空閒媛衝槽1 0 4 ,如單複製傳 輸一樣地傳輸訊框及回收孃衝器,如果沒有被找到,媛衝 器管理員72傳輸多複製訊框並放置一個複製號碼為“-1” 的訊框指標的複製品到多複製佇列90。當主機用完一個排 列在擴充匯流排輸出佇列75或管理埠輸出佇列77(經PCI介 面26)的多複製訊框之訊框指標後,主機會將一個複製號 碼為“-1”的訊框指標副本,經訊框指標暫存器寫入多複 製佇列中。此暫存器是第2圖暫存器60方塊圖中所描述的 —個暫存器。 類似輸出佇列74,多複製佇列90構造成輸入路徑及輸 出路徑。輪入路徑或稱為書寫端,允許璋向量FIFO 70及 鍰衝器管理員放置訊框指檷/複製號碼到複製佇列90。輸 入路徑或稱為讀取端,允許多複製佇列90放置訊框指標到 多複製快取96。提供更多訊框指標/複製號碼儲存容量的 記憶區稱為多複製佇列過載區124,此一多複製佇列過載 區由外部記憶體36提供。 __ 當訊框指檷/複製號碼被寫入一個多複製佇列90中時 尺度逋用中關家科{ CNS ) A4坑格(210X297公廣) '~~ 31 91472 --------W裝-----^--訂------束 (請先M'讀背面之注意事項再填艿本頁) 804530 A7 ii7 五、發明説明(32 止 為 Τ 滿 4 9 端 取 請 到 直 的 ο 9 列 貯 製 複 多 入 寫 要 4,碼 9 虑 tl 端a * 複 0/ 到標 過 指 傳框 2 L, 9SP 端的 寫 多 書 更 由有 們如 它 C- 2 區 載 過 的12 36區 體載 憶過 記 及 部94標 外端指 於取框 位 _ 訊 ΙΛ^ ny ^7 ft ft 寫90多 會 列過 將ff中 製 列 複f? 端 多 製 寫0複 書 一 多 中 人端 放寫 r 書 了人 滿 《 都 姶 4 5 2 開 丨碼 虎 製 複 著 持 維 被 , 序後 次置 的位 90出 列除 佇清 4 $ 9 複端 多取 過讀 通的 列 佇 標 指 框 訊 製 複 多 要 只 得 使 由 將 碼 號 製 複 端 12寫 區 書 載的 過列 列佇 佇製 製複 複多 多由 端 取 讀 的 ο 9 列 佇 製 複 多 至 移 及 (諳先阽讀背而之注意事項再填寫本Η ) '裝. 區 載 過 列 庁 製 複 多 rt- 至 移 取 快 πυ 歹 佇 製 複 多 個 1 供 提 但 ο 9 列 佇 製 複 多 Μ 類 取 快 列 佇 製 複 多 碼 號 複 / 標 指 框 訊 描 掃 供 Μ 區 尋 搜 可 列 佇 製 複 多 由 2 7 員 j 理 管 器 衝 媛 C 料 資 人 輸 個 6 5 2 有 持 碼 號 製 複 據 根 標 指 框 訊 取 讀 製 複 多 入 放 或 訂 標 指 框 訊 此 理 處 或 6 9 取 快 7ί簞 端果 取如 , 的輸 74傳 列程 佇排 出會 5 2 輸 7 由 員 2 I 7 理 員 管 理器 管衝 器緩 衝 , 媛 標 次指 每框 * 訊 者 個 再 一 取 請 M沪部中戎標苹扃貝工消费合作社印聚 理訊 管的 器 ” 衝„1 緩纟 為 ),碼 框號 訊製 製複 複有 多否 個 是 1 看 示 , 6 表 9 ί列 ’佇 „ο製 “ 複 為多 元 描 位掃 製72 複員 員 ” 1 理 “ 管 為器 爵 B 0 镄 號鑀 製 . 複了 , Ϊ 標找 指果 框如 傳 次 一 後 最 框 是 這 示 表 及 料 資 入 輸 勒 移 間 期 輸 傳 在 員 理 管 器 緩 到 I 找“ 有碼 沒號 果製 如複 〇 個 槽 一 衝 置 m放 間 , 空 後 。 到之 9 器成列 衝 完 佇 媛輸製 谡 傳複 歸在多 至 標 指 框 訊 的 本紙張尺度適用中國國家椋準(CNS ) Λ4規格(2】ΟΧ 297公釐) 32 9 1472 <? s 〇8Q a7 B7 經Μ-部中央標準局資工消费合作社印裝 五、發明説明(3 3 ) 1 1 _衝器管m員72經由讓取訊框指標/ 複 製 號 碼 來 腋 務 1 i 多 複製ff列9 0 t 將訊榧指標/複 製 號 碼 放 入 多 複 製 快 取 y 6 1 1 j 中 .或者處理訊框指標/複 製 號 碼 〇 這 m 動 作 跟 訊 框 傳 輸 請 先 Μ. 讀 背 而 之 注 意 事 項 再 1 無 關。K 「'.舉 1兩個例子則跟媛衝器管理員_取訊框指標 Ί 的 1 ) 複製號碼有關,看複製號碼是“ 1”或"-1 如果鏈衝器管理R 7 2由多複製佇列9 0讓取- 〇 -涸複製號碼 1 1 I 1 1 ** > 1 " 的 訊 框 指 標 〇 假設 此 時 多 複 製 快 取 96有 儲 存 空 間 1 裝 | ,它 將 __. 個 新 的 輸 入資 料 寫 入 〇 如 果 多 複 製 快 取96沒有 % 本 η 儲存 空 間 , m 衝 器 管理 72必 須 清 除 多 複 製 快 取 96的 1 1 一也 空 間 〇 清 除 空 間的 做 法 是 經 由 取 多 複 製 快 取 96 1 1 | 中較 舊 的 訊 框 指 標 /複製號碼, 用該多複製快取96中的 1 複製 號 碼 更 新 在 外 部記 憶 njn 體 36中 的 訊 框 指 標 鍰 衝 器 表 訂 I m , m ι·>·>\ 後 刪 除 此 在 多複 製 快 取 96中 較 的 輸 入 資 料 ♦ I 1 如此 即 可 清 除 出 一 個儲 存 t/tr 間 0 一 巨 有 了 儲 存 空 間 1 1 1 新的 訊 框 指 標 /複製號碼就可以寫人多複製快取96 了。 1 2 )如果緩衝器管理R 7 2由多複製佇列9 0譲取- -個複製號碼 丨 1 ^ _ | " 的 訊 m 指 標 〇 則Μ 衝 器 管 理 員 72搜 尋 多 複 m 快 取 1 1 I 96 , 看 是 否 有 訊 框 指標 複 製 號 碼 a > = 1的訊框指標。 Μ 1 ί 下兩 個 案 例 t 取 決 於媛 衝 器 管 理 員 72是 否 在 多 複 製 快 1 1 取96找 到 符 合 條 件 的訊 框 指 標 0 ί a )如 果 m 衝 器 管 理 員72找到 個 符 合 條 件 訊 框 指 標 , 1 * 假 設 多 複 製 快 取 96的 輸 入 資 料 複 製 號 碼 為 a 1 ” , 1 I 緩 衝 器 管 il 員 7 2會刪 除 多 複 製 快 取 的 輸 人 資 料 t 並 } 1 I 將 該 訊 框 指 摺 放 人闸 收 佇 列 98 〇 如 果 多 複 m 快 取 96 1 1 本紙張尺度適用中國國家榇準(CNS ) Λ4現格(210X297公釐) 33 91472 <? ο Θ Ο Α7 Β7 經濟部中决標準局負.1消费合作杜印¾ 五、發明説明( 3 4 ) 1 I 的輸人資料複製號碼為 4 > 1 ” 媛 衝 器 管 理 員 72會 1 I 將該複製號碼減- ~ 〇 ί 1 b )) 縛 衝 器 管 理 72沒 有 找 到 個 符 合 的 訊 框 指 標 , ίV 請 先 • I 1 1 這 表 示 符 合 的 訊框 指 m 已 經 移 到 外 部 記 憶 體 36上 閱· 讀 背 1 的 訊 框 m 結 串 列的 嫌 衝 器 表 _ 〇 m 衝 器 管 理 72 而 I 之 1 vi | 必 須 跑 到 此 緩 衝器 表 頭 並 謓 取 此 複 製 號 碼 〇 如 果 意 事 項 1 I 此 複 製 號 碼 的 值為 1 ” 則此訊框不再需要, 再 填 1 嫒 衝 器 管 理 員 72將 此 訊 框 指 標 放 入 回 收 佇 列 98中 寫 本 装 頁 1 r> 如 果 此 複 製 號碼 的 值 (在記憶體中)為 a > 1 t ! I m 衝 器 管 理 員 72會 將 此 訊 框 指 標 /複製號碼的副 1 1 I 本 放 入 多 複 製 快取 9 6 9 並 將 複 製 號 碼 減 一 〇 如 果 1 訂 多 複 製 快 取 96已滿 t 緩 衡 器 管 理 員 72移 動 —1 個 較 1 舊 的 訊 框 指 標 /複製號碼到外部記憶體3 6 , Μ便 1 1 清 出 —. 個 儲 存 空間 0 1 1 回收 佇 列 98儲 存 有 指到 不 需 要 的 鏈 结 串 列 的 訊 框 指 標 1 0 賴 衝器 管 理 員 7 2當 它 服務 多 複 製 快 取 1 發 現 訊 榧 指 標 的 1 i 複 製 號碼 的 值 為 U Γ, > 它會寫- -値訊柢指標到回收佇列 1 1 I 〇 (此即訊框的最後傳輸已成功地達成) 0 再 者 y 埠 向 量 1 1 FIFOS T 列 的 情 況 下 ► 也會 寫 訊 框 指 標 到 再 漀 明 佇 列 98 1 1 : 1 ) 訊框指標的埠向虽是空的, 或2 ) 因 為 所 有 的 埠 向 量 '| 轉 送 佇列 都 滿 了 i 導 致 訊框 指 標 無 法 排 列 0 最 後 當 主 m 41 使 用 完一 個 由 擴 充 匯 流 排埠 輸 出 佇 列 77 或 管 理 埠 輸 出 佇 1 列 75排列 的 單 複 製 訊 框 後, 主 機 會 寫 一 個 訊 框 指 標 到 回 收 1 佇 列 9 8 ί使用訊框指搮暫存器) 0 1 1 本紙張尺度適用十國國家標皁(CNS ) A4規格(2!0X297公釐) 34 9 1472 、3 0 8 Λ7 H7 經濟部中央標準局員工消费合作社印製 五、發明説明( 3 5 ) 1 1 當 緩 衝 器 管 理 72處 理 回 收 佇 列 輸 入 資料 ,它 運作一 1 俩 r;R 枢 指 的 m 结 1 Μ 便 歸 m 每 個 摄衝 器到 空閒緩 1 衝 槽 1 0 本發明的範例實胞例中, 並沒有描述回收佇列 請 1 1 ) 的 内 部 結 構 t 以 本 發 明 的 m m 實 施 例 而 言 ,回 收泞 列只含 先 閱— 讀 Ί I 有 訊 框 指 標 (1 4位元) 的 結 構 〇 回 收 佇 列 的 書寫 端保 持有64 背 1¾ 之 1 1 個 輸 入 資 料 i 回 收 庁 列 的 m 取 端 保 持 則 有 1 6個 輸人 資料。 注 意 事 1 1 類 似 輸 出 佇 列 7 4 ♦ 回 收 佇 列 98構 造 成 輸 人 路徑 及輸 出路徑 項 再 1 〇 輸 入 路 徑 或 稱 為 書 寫 端 ♦ 允 許 媛 衝 器 管 理員 72放 置訊框 % 本 Τι 裝 I 指 僳 到 回 收 庁 列 98 〇 輸 出 m 徑 或 稱 為 謓 取 jLiir 端t 允許 媛衝器 1 I 管 理 員 72_ 取 訊 框 指 檑 及 歸 湛 所 有 相 關 的 媛衝 器到 空閲緩 1 1 1 衝 槽 10 4 . 更多訊框指標儲存容量由回收佇列過載區1 2 2提 1 1 訂 供 * 此 回 收 佇 列 過 載 區 係 由 外 部 記 憶 體 36所提 供。 1 當 訊 框 指 標 寫 入 —« 個 空 的 回 收 佇 列 98中時 ,它 們由書 1 i 窝 端 100傳遞到讓取端102 , 直 到 m 取 端 102滿了為止。如 1 1 有 更 多 的 訊 框 指 標 要 寫 入 回 收 佇 列 98的 書 寫端 100, 將會 1 寫 人 位 於 外 部 記 憶 體 36的 過 載 區 122 中。 - S回收佇列98 1 I 的 讚 取 端 102及過載區122都 滿 了 , 放 入 回 收佇 列98中過多 1 1 | 的 訊 框 指 標 開 始 填 人 書 寫 端 1 00 〇 1 1 第 1 1圖 描 述 空 聞 鑀 衝 槽 104的内部结構的- -個範例實 1 1 胞 例 , 空 間 緩 衝 槽 104 是- -個 F IF0, 内含- -個位址指標指 i 到 外 部 記 憶 體 36中 所 有 的 空 間 媛 衝 器 1 40 〇 當所有的訊框 J ί 部 被 收 到 了 * 緩 衝 管 理 員 72由 間 緩 衝 槽 10 4捕捉可用的 1 I 位 址 指 標 Η 儲 存 新 進 來 的 資 料 〇 媛 衝 管 理 員72也由 空間媛 1 I 衝 槽 1 0 4配置位址指撢到主機處理器2 8 (當 被要 求如 此做的 1 1 本紙張尺度適用中國國家標準(CNS ) A4規格(2】0X297公釐) 35 9 1472 4 0 A7 B7 經漭部中央標準局員工消费合作社印发 五、發明説明(36) 時候)。ΐ機可經由寫或_取位於直接輸入或輸出空間内 的(cM r e c「 i n f> ti t / ο ϋ t p ii t s p a c e )空閒緩衝暫存器6 0中的 一個暫存器,而要求或歸谡位址指標到空閒緩衝槽1 〇 4。 在本發明書寫端106及空閒鍰銜槽104的讀取端108都分別 持有著64個輸人資料。 空閭媛衝槽1 0 4構造成一個輸人路徑及一個輸出路徑 。(類似輸出佇列74),輸入路徑或稱為書寫端106,允許 鍰衝器管理員7 2或主機2 8放置位址指標到空間媛衝槽Π) 4 。輸出路徑或稱為謓取端108,允許媛衝器管理員72提供 位址指標到主機2 8,或由空間缓衝槽1 0 4拉出位址指標Μ 供儲存接收訊框資料。儲存更多可用位址指標的儲存空間 ,由外部記憶體36的空閒媛衝槽過載區120提供,如前所 述。 交換裝置開始動作的時候,空間嬢衝槽由讀取端 108產生位址指標ρ當訊框來臨,在空閒緩衝槽1〇4的空閒 表列被謓取。如果在書寫端106沒有足夠的鍰衝器指標來 控制資料通訊,過載區120被存取Κ得到更多的嫒衝器指 標。 在交換裝置12開始動作的時候,本發明的一個特定範 例實施例·提出較哇的嫒衝器指標提供的方法。當交換裝 置第一次接上電源時,在外部記憶體36中的過載區120並 不需要有嫒衝器指標。反而,緩衝器指標是在蓮轉期間產 生的。交換裝置12接上電源時,可產生緩衝器指標,並可 將其放人過載區]2 0中,但是,此時有1 6 0 0 0到3 2 0 0 0個這 本紙悵尺度適汛中國國家標率(CNS ),\衫見格(210Χ297公趋) 36 91472 — — --------# ! (請先閱绩背vg之注意事項再填3本頁) < 經泸部中央螵準局負二消費合作社印絮 五、發明説明 ( 3 7 ) 樣 的 婿 衝 器 指 標 1 這 樣 食 m 低 交 換 裝 賈 12開 機的 動 作 。本 發 明 採 用 了 -一 ijb 開 間 期 間 奸 的 特 點 例 如, 所有 的 m 衝器 都 是 空 間 的 這 媛 衝 器 的 識 別 碼 也 都 被知 嘵等 η 因 此, 在 開 槠 之 後 當 需 要 用 到 這 些 緩 衝 器 的 B寺候 .再 使 用 計數 器 1 8 0去產生這些鍰衝器。 有如第1 0圖描述的一樣。 空閒 計 數 表 列 產 生 器 18 0 ( f r e e 1 i s t C 0 u η t gene r a t 〇 Γ ) 被 連接 到 多 工 器 1 8 2的輸人。 茌開機的時候, 因為在空閒媛衝槽 10 4的空閒表列是空的, 所以空間訐數表列產生器1 80 產生 m 衝 器 指 標 0 一 旦 空 閒 表 列 達 到 它 的 最 高計 數時 V 它 就不 再 產 生 緩 衝 器 指 標 〇 當 一 個 訊 框 封 包 被 交 換 裝 置 1 2接 收 後, 該訊 框 封 包會 被 打 散 到 固 定 長 度 的 緩 衝 器 中 * 般 而 言, 訊框 的 大 小是 會 變 動 的 〇 該 些 m 衝 器 有 256個位元組長,存放資料的部 分 則 有 2 40個位元組長。 在傳輸緩衝器的内容之後, 媛衝 器 指 標 被 放 人 回 收 佇 列 98 中 如 杲 緩 衝 器鐽 (b u f f e Γ c h a i η ) 能 夠 運 作 * 則 直 接 放 到 空 鍰 衝 槽 1 04 〇 在交換裝置 1 2的 操 作 期 間 任 何 回 到 空 鑀 衝 槽 10 4的位址指標, 由書 寫 端 106傳遞到讓取端1 0 8 t 如 果 謓 取 端 108滿了, 多的位 址 指 標 將 會 寫 入 位 於 外 部 記 憶 頒 36的 過 載區 1 2 0 _ 3 0 -旦 讀 取 端 10 8及過載區1 2 2 都 滿 了 1 再 放 人 空閒 猨衝 槽 10 4的 位 址 指 標 , 則 將 開 始 填 入 空 間 媛 衝 槽 10 4的書寫端1 0 6 0 第 13 圖 是 根 m 本 發 明 —1 實 施 例 描 述多 複製 佇 列 9 6的 內 部 管 理 方 塊 表 示 阚 〇 如 前 所 面 的 簡 單 討論 ,進 人 多 複製 快 取 9 6 輸 入 資 料 的 時 間 先 後 的 次 序 被 保 持著 。在 本 發 明中 本纸張尺度適用中囷國家榇率(CNS ) Λ4規格(2丨OX 297公釐) 37 91472 53〇8〇 A7 B7 Μ"•部中央標準局貝工消费合竹社印欠 五、發明説明( 3 8 ) » 如 -W-t. m 述 技 U > 時 間 先 後 的 次 序 並 非 由 時 間 戳 記 所 保 持 而 是 m 12 憶 體 的 茛 體 次 tie t · ( P h y 3 1 C a Ϊ 0 r d e r i η S ) m保持 η 本 發 明 的 多 複 製 快 取 9 6 Ιθ] 時 必 須 m 免 使 用 正 確 性 位 元 , 反 而 是 使 用 解 碼 正 確 性 這 — 方 面 將 在 以 下 Μ 論 〇 參 考 到 第 13 圖 » 多 複 製 快 取 96被 構 造 成 . 個 四 通 道 的 1 各 組 相 關 的 (S e t —a s S 0 c i a t i v e )記憶體, 任何進人多複 製 佇 列 9 6 的 輸 人 資 料 * 包 含 訊 框 指 標 及 其 複 製 號 碼 t 如 前 所 述 〇 訊 框 指 標 的 最 低 個 位 元 決 定 要 將 訊 框 存 人 各 組 相 關 快 取 96 中 的 哪 — 行 * 在 本 發 明 —1 實 施 例 中 雖 然 可 把 快 取 的 尺 寸 變 大 , 且 它 的 行 數 並 沒 有 限 制 ♦ 但 本 發 明 的 快 取 96 只 有 6 4 行 〇 各 m 相 關 快 取 96將 所 有 的 行 數 再 分 成 四 大 行 > 每 一 大 行 都 可 並 行 地 搜 尋 〇 當 緩 衝 管 理 員 72 把 個 輸 入 資 料 存 入 快 取 9 6 中 時 r 它 永 遠 存 人 最 左 上 的 位 元 (5 1 - 39 ), 此係由 訊 框 指 標 最 低 6位元所指定的。 當此大行被譲取後, 全部 的 輸 入 資 料 向 右 移 13 個 位 元 ? 且 此 — 大 行 將 被 再 寫 回 〇 被 實 際 寫 入 快 取 96 中 的 輸 入 資 料 * 包 括 此 訊 框 指 標 最 左 8位 元 的 位 址 標 m (a dd re S S t a g ) » 及 5位元眼此訊框指標有 關 的 複 製 號 碼 〇 當 此 輸 入 資 料 被 謓 出 快 取 96 後 t 訊 框 指 »τιττ 標 内 訊 框 標 籤 及 標 示 快 取 96 内 行 號 的 所 有 位 元 將 被 重 新 整 理 □ 如 果 快 取 滿 了 t 而 又 有 新 的 輸 人 資 料 要 寫 人 , 快 取 中 最 η 的 輸 人 資 料 將 由 快 取 中 移 走 〇 如 前 所 述 有 關 於 m 衝 器 1 4 2的情形, 专走訊框指標的複製號碼將被寫入外部 本紙張尺度適Λ中國國家標準(CNS ) Λ4規格(210X297公釐) 38 91472 53〇8〇 Α7 R7 五、發明説明(3 9) 紀情聘中被移走的訊框指標所指到的訊抿嬢衝器表頭1 4 2 。因此,訊框儲存於外部記憶體3 S中(即緩衝器1 4 0 ),此 外部記憶體檐任多複製快取9 6的過載區,用Μ儲存複製號 碼。 本發明一個優越的特點是在各組相關快取9 6中沒有分 離的正確位元,當複製號碼是0 0 0 0 0時,緩衝器管理員72 知道這個輸入資料不再ίΕ確,所Κ將此輸人資料由快取96 中移走。如此可簡化快取的结構。另外本發明快取9R的其 他優點是它允許執行快速地搜尋。因為鍰衝器管理員7 2只 需檢査一行,而此行已由離開多複製佇列90的訊框指標所 決定。四大行的輸入資料被並行的檢査,更進一步增加搜 尋的速度。雖然只如同一個四通道各組相關的記憶體來敘 述本快取結構,但這只是一個範例,用來表示一個η通道 各組相關的記憶體並未脫離本發明的範圍。 由以上的敘述,我們可Μ知道本發明可經由擺放輸入 資料在各行中的實體位置而保持此輸人資料的時間次序( ^¾部中央標隼局員工消費合作社印β (讀先閲讀背兩之注意事項再填巧本頁) 輸入資料的新舊),換句話說,此輸入資料的實體位置表 示此輸人資料的新舊程度。要變舊一個輸入資料可由重新 整理此記憶體來達成。 一個本發明特定的實施洌,Μ —埠接著一埠的基礎, 可訂製由交換裝置12切換的訊框等待時間(the latency of frames)。參考第14刪,埠向最FIFO 70檢查接收埠的交換 裝置模式(programmed switch mode)(此交換裝置椹式是 可改變的),Μ決定甚麽時候要將訊框指標及相關資訊放 本紙張尺度遍用中國國家標丰(CNS ) Λ4规格(2丨0'〆297公釐) 3 9 9 1 472 4 53 0 8 ϋ Α7 ί£"部中央樣本局員工消费合作社印製 Β7五、發明説明(4〇) 入適當的傳輸埠的輸出佇列74中。第一模式(等待時間質 模式)埤向最卜' I F 0 7 0沒有限制甚麼時間放訊抿指標到輸出 埠7 4 n第二m式(中等待時間橫式)埠向i F I F 0 7 0放置 訊框指標到輸出佇列7 4 ,只在該訊椐指標的6 4個位元被接 收到之後.第三m式(高等待時間模式)埠向虽P I F ϋ 70 放置訊框指標到輸出庁列7 4中,Κ茌該訊框指標全部被 接收之後。 這裡有一些特殊的例子,必須改變埠向贵F I F 0 7 0傳遞 訊框指標到輸出佇列? 4的時間。1 )由第一模式或第二模式 的1 0 M b / s埠轉送訊框到1 0 0 K b / s埠。2 )訊框轉送到管理埠 3 0。3 )訊框轉送到擴充匯流排埠。在1 )中,1 0 M b / s埠與 1 0 0 M b / s埠速率的不同,使得轉送模式強制變到第三的高 等待時間模式。在2 )中.所有的訊框傳遞到管理埠使用的 是第三模式的訊框。在3 )中,所Μ傳逓到擴充匯流排埠的 訊框使用擴充匯流排埠26的切換楔式。當一個多複製埠向 量包含有這些特殊的埠中一種時,就必須使用最長的等待 時間切換模式。例如,有一個訊框由第一或第二構式的埠 所接收,但它的多複製埠向量包含有管理埠3 0 ,於是切換 模式必須使用第三種横式。在這種情況下,只有訊框指標 全部被接收之後,一個訊框指搮的副本再放到輸出貯列7 4 中0 交換裝置横式將做更詳細的討論,施加到輸人埠的交 換裝置模式可決定轉送等待時間(交換裝置1 2 —旦接到訊 抿後將Κ多快的速度轉送訊框 >,及可決定降低到每一輸 (誚先間讀背而之注意事項再填巧本頁) 本紙張尺度適用I丨’國國家標準(CNS)A4規格( 210X 297公釐) 40 9 1472 45308; A7 經消部中央標羋局員工消费合作社印" B7五、發明説明(4 1) 出增的Η段/錯誤fragment/error· propagation)延遲的能 力„再者,第二憤式,即中等待時間橫式是每一個埠的預 設模式。而在暫存器6 0中,可以以一埠為頭位來規副交換 裝置横式。 在三個交換裝置_式中,由内部M A C埠F 0 5 2接收 的訊框,將儘可能轉送到外部記憶體緩衝器1 4 0中。大約 同一時間,規則檢査器4 2或5 8 ,接收目的位址及來源位址 -接收埠向量,框訊指標及一些附加訊息,然後執行適當 的檢査„ 一目.檢査完成,規則檢查器42或58歸遒訊框指標 及轉送埠向量到F I F 0 7 0。 埠向最 FIFO 70將訊框指標放到輸出佇列74的書寫 端 76,提供埠向景Μ識別輸出埠。當埠向量FIFO 70接收 埠向量(訊框指標)及將訊框指標放到輸出佇列74這一段時 間,接收埠的交換裝置模式定義此埠的轉送等待時間。這 是為下列三棰模式所做的一個敘述。一旦訊框指標傳遞到 輸出佇列7 4的讀取端7 8 ,緩衝器管理員7 2 _取訊框指標並 排程傳輸的時間。緩衝器管理員72開始由訊框指標指到的 位址移動訊框資料。一旦傳輸MAC埠的FIF054回到它最先 開始的點(假設媒體可用於傳輸訊框資料),則開始傳輸訊 框。 第一模式被設計來提供一最低等待時間的傳輸模式, 訊框的傳輸及轉送Μ線速率的速度(1 i ti e - r a t e d s p e e d )來 傳輸,m第一m式而言,因為訊框在尚未確定此訊框為一 Η段(即小於2 R位元)或有包含C R C错誤之前,即排列Κ供 本紙张尺度適用中國國家標準{ CNS ) Λ4規格(2】0:<297公釐) 41 91472 (誚先虮讀背而之注意事項再垆'"本頁) 4 53 0 A7 R7 經浐部中央標卑局員工消费合作社印裝 五、發明説明(4 2 ) 傳輸,所W沒有何的網K錯_保護(n e t w 〇 r k e r r 〇 r P「o t e c t; ί ο n )。第一模式中,輸出埠的訊框傳輸之前,訊 框的接收可能尚未完成。如果訊框的接收结束於一個不成 熟的傳輸週期(r u n t ),或帶有一個不正確的C R C ,接收M A C 會標示< m a「k >外部記憶體3 6的緩衝器1 4 2 , K顯示目前發 生這些情形。傅輸m a c保證如果開始傳輸一訊框,而此訊 框隨後會結束於一個不成熟的傅輸週期.或有一錯誤的 C R C ,則Μ Λ C會產生一個壞的C H C。M A C尚未開始訊框的傳輸 ,但鍰衝器142已標示訊框結束於一個不成熟的傳輸週期 或有一錯誤的C R C,緩衝器管理員將不會轉送訊框到輸出 埠。 第二横式提供一個低等待時間訊框轉送横式,但具有 些許的網路錯誤保護。在接收到64位元或者更多位元之後 ,訊框才會轉送。此會允許交換裝置12過滤掉(即不轉送) 片段的訊框。然而,此模式無法完全過瀘位元數大於64、 有CRC錯誤的訊框。 第二横式,達到B 4位元臨界值之訊框的訊框指標排列 到適當的輸出佇列7 4。未達到6 4位元臨界值之訊框將被刪 除,其訊柢指標也不會放人輸出佇列7 4。如果一個大於或 等於(5 4位元的訊框结束於一個不正確的C R C ,接收M A C標 示(m a r k )外部記憶體3 (3的緩衝器1 4 2 , Μ顯示目前發生這 種情形。如果傳輸開始於一個大於或等於64位元的訊框. 而隨後结束於一個不正確的C 1? C ,傳輸Μ Λ C會以一個壞的 C K C結束此傅輸如果一 flf傳輸H /\ C舱未開始傳輸訊框,但 ---------装--------訂 .(誚先閱讀背面之注意事項再填」巧本頁) 本紙張尺度適用中國闼家標準(CNS ) Λ4規格(210X297公釐) 42 9 1472 4 53 ΟThe Ministry of Economic Affairs and the Central Bureau of Standards of the People's Republic of China negative T_consumer cooperation Du Yinzhuang 5. Description of the invention (26) 1 1 standard 〇 4) Update the M 1 B counter. 1 The buffer manager copies a specified frame to all the appropriate outputs. 1 1 After 7 4 75, the port direction is up to KIP 0 7 0 si- Count the number of copies (copy number) * 1 I M and place the signal. Index and copy number to the multi-copy queue 9 0 of the writing end 92 First read the + on the back + 丨 I η If the copy number is zero *, it means that the frame is not transferable. If it is one I 1, then 1 means yes — a single copy transmission If the number is greater than the fee t, it means 1 1 yes—1 multi-copy transmission t These three examples will be explained below 0. When re-filled 1 ά I indicates that the frame indicator has a zero transfer Port vector 9 and no bit has any bit set f Port vector F X F 0 7 0 Directly pass the frame pointer to the writing end of the recovery queue 9 8 at 1 1 I 10 When the port vector FIF070 service Recycle queue I 1 1 98 hours »This situation will be described in the next 0 punch The manager 72 interrupts the sequence of the yuan punch r 1 > assigns the address index of each fre e buffer to the order 1 free m and the 104 writing end 1 06 〇1 1 When the copy number is watched * indicates a single copy transmission »port direction 1 1 up to FIF 0 7 0 copy frame indicator» control signal / control opcode, and VLAH 1, ice indicator to the appropriate output queue 7 4 The port vector F IF 07 0 sets the generation copy bit in the output queue I 1 7 4 (see Figures 5 and 6). K indicates that this is a single 1 1 I copy transmission when the buffer manager 72 is sent by the port Output queue 74 _ Frame 1 i When the indicator and the generation copy bit * m The punch manager 72 will do the scheduling of the I 1 transmission as described previously. The buffer manager 72 uses the frame indicator to configure the External memory: 1 The first buffer in body 3 6 r This element stores information Box 〇 Buffer manager 72 please take the first-* buffer to get the data from the first buffer, and I send the data to the appropriate MAC transmission F IF0 54 〇 Suppose the frame expands to multiple 1 i buffers * The above-mentioned m knot will m knot to the next buffer f so 1 1 This paper size applies to the Chinese National Standard (CNS) Λ4 specification (210 × 297 mm) 26 9 1472 ό B7 Central Bureau of Standards, Ministry of Economic Affairs Consumers ’Cooperatives, Fanfan V. Invention Description (2 7) The structure provides m punch manager 7 2 — addresses * can be used to find and transmit all buffers that are critical to the frame 0 once the data is placed After the FIFO 54M is available for transmission, the buffer will be invalidated and returned to the free m slot 10 4, and the M stem will be reconfigured to provide other frames to store data. When the complex number m is larger than the port number * Port direction KI F0 7 0 Copy the frame indicator, control signal / control opcode, and VLAN indicator to the appropriate port output queue 7 4 〇 (This is referred to queue 7 4 In the case of queues 5 and 77, the same is true.) Port 0 clears f I FQ7 0 to clear the output of single copy bit t in the appropriate frame indicator in queue 74 to copy the number greater than one. Put the frame pointer into the copying end of the writing queue 90 2 2 〇 P Ask the buffer manager 72 to read one of the frame pointers and the output queue 74 to clear the list—f Yuan Chong when copying the bits The server administrator will schedule the transmission of the frame. At the same time, check the multi-copy cache 96 * to see if the frame index of the input data has a copy number of 1. If a copy number of 1 is found in the multi-copy cache 96 Box indicators. Yuan Miao Administrator 72 with single-frame m transmission of the frame—1 sample * frame scheduled for round transmission * and the m frame is recovered during transmission. If the frame index is not in multiple m In the cache 9 6 or the frame index in the multi-replication cache 96 is greater than 1, the element manager 72 will transmit the frame but does not reclaim the buffer. After the transmission is successful, the element manager 72 places A copy of the frame indicator > With the copy number "-1? To the multi-copy queue 90 output 92 0 When each multi-copy frame pass is passed * as long as the buffer manager 72 is not in multi-copy The cache rule in the cache 96 has a frame indicator with a copy number of 1, and the m puncher tube m member 7 2 will place a copy of the frame indicator to multiple copies. The secret standard is the general Chinese national standard (peak (do n’t x gift) (Chu) 2? 91472 ° S〇B7 Central Standard of the Ministry of Economic Affairs :: f- Bureau of the Ministry of Work and Consumers Cooperative Fifth Edition, Invention Description (2 8) 1 1 in column 9C so t at any time * The multi-copy storage column 90 will contain the duplicate Ί number.) 1 ί The internal frame indicator or the frame indicator that requires multiple copy numbers-1 The frame indicator 2 This second reading! 1 m punch manager 7 2 often serves multiple replication queues 90 1 and multi-copy fast reading I back 1 takes 96 K and m is received as m's punch when it is multi-copying ning 90 f reading surface 1 note 1 Take the message with the complex m code * 1 > 1 ”means that the instrument manager will try this one item | new rounder information (frame indicator and copy number) into the multi-copy cache 96 〇 For example, J / i will copy the cache 96 full * Buffer Manager 72 will leave space for the new price. This page contains 1 box indicator. Buffer Manager 72 謓 Get an older input data »Update at 1 1 External memory 36 Copy number of the input data in the header of the yuan puncher »1 \ Then multi-copy queue 96 to clear this older input data. 〇 If there is more space in the multi-copy [1 custom cache 96, you can use the engine buffer The device manager 7 2 can send new input 1 1 input data from the multi-copy queue 90 1 to the multi-copy cache 96 0!! Buffer manager 72 serves the multi-copy queue 90 * and reads the copy number 1 1 -1 frame index t It searches the multi-copy cache 96 4 for the frame index address of the copy number ί, VJC 1 ”. M is easy to reduce or delete. If the buffer 1 1 manager 7 2 sends a signal index of U > =, the buffer management # 72 1 I will 1) If the copy number is greater than 1 r, reduce the frame index of the multiple copy cache 1 1 1 * Or 2) If the copying Dk & code is exactly 1, delete multiple copying t) ί Pick 'medium 1 1' This frame refers to m / copy several input data, so i do n’t put the frame index into the recovery 伫: 丨 1 Column 98 〇1 If the m punch manager does not find the copy number t * > = 1 ', the frame 1 1 pointer. The punch manager 72 will search for the frame pointer in the external memory 36! | M punch Device header (see Section 9), look for JJ ί «3 conforming copy number nitrate 0 if it is suitable for 1 paper size + National Standard (CNS) Λ 4 grid (2) 0X297 mm 28 91472 3〇 〇A7 B7 Carp " Printed by the Peasant Poverty Alleviation Co., Ltd. of the Central Standards Bureau of the People's Republic of China 5. Invention Description (2 9) 1 | The copy number in the memory is * 4 1 " η Buffer Manager 72 will put the frame index 1 1 mark into the recycling f? Column 98 0 if the number in memory r | W. 屮 is "> 1 〇 ii punch tube m 72 place the frame pointer and copy the number to Replicate m queue 96 * Please first 1 1 i and then reduce the copy code 〇 Read 1 m punch management 72 often M Please take the frame instructions to serve the back 1 1) Note 1 close the list 98 〇 then operate (ν a I king) The link string can be buffered to empty f and refilled 1 I free m punch slot 10 4 〇 This action is only returned to the punch 1 which has a zero port to the frame, and From the port to the most FIF 0 7 0 line up to the recovery queue 0 or there are multiple copies of %% installed on this page 1 system to the most and all transmitted frames 0 as described above when the single copy message i I frame is After the transmission t single copy of the message buffer It will be directly assigned to m 1 ί I to the buffer slot 104〇1. If the output queue 74 and a M.m load area 11 0 1 in the external memory 3 6 are full, the port buffer FIFO 7 0 cannot Put the frame indicator with the single copy transfer vector 1 1 into the output queue 74 r The frame will be discarded 0 The m frame! 1 indicator will be returned to m to the recovery queue 98 * The filr action of the discarded frame is parented Device J management data (man age ϊη e η t re SO υ Γ ce 0 fthe S w 1 t C h) recorded Ί II 〇 If one or more output queues and their external memory 36 I The overload area 1 1 in 1 1 is full. Port vector FIF 0 70 cannot be placed—1 or more frame indicators 1 with 1 1 multi-copy transfer vector 1 So the frame can only be transferred to m with 1 1 space Queue and repeat m number into multiple copy queue 90 • Only 1 alone reflects completed The frame index 0 that is placed successfully can not be placed (η on -P 1 ace me nt 1 I). The frame index will be used by the frame pointer to indicate that the switch device management that cannot arrange the port 1 1 I resources will be recorded if. Or more The output queues M and their 1 ± external 1 1 paper standards are common in China_Home Standards (CNS) M Qing 10x ™ 2g 7 2 〇8 0 Α7 Β7 ^ V. Description of the invention (30) The overload area 110 in the memory 36 is full, and one or more of the frame index cannot be placed in the port FIP070. The frame index will be passed to the recycle queue 9 8. The switching device management resources are formally notified. Port direction MF 0 7 0 uses the editor that can complete the above actions. For example, to determine which output queue is full, only put the frame pointer into the queue that is not full, and the other ones that are full are discarded. Such logic can be easily provided by the basic skills of the technical transformation described in this function. The output queue is full and moss, which can be determined by comparing the number of frame indicators of a particular output queue (this value is recorded in a register 60), and a threshold value indicating that the queue is full compare with. Based on the cognition on KK, for multi-copy frames, the port direction most FIF 0 0 0 only discards those frame indicators that will be put into the output queue, and increases the value of the discarded frame (save in other temporary frames) Register 60). The multi-copy queue 90 is a high-priority queue used by the buffer manager 72. Before all buffers (ie, address indicators) store frames, it is used to record how many specific multi-copy frames need to be transmitted. To accomplish this, all the punches can be recycled to the free buffer tank 104. The writing terminal 92 and the access terminal 94 of this output queue respectively hold 64 and 16 input data. Multi-replication queue 90 feeds data to multi-replication cache 96, and multi-replication cache 96 is used by buffer manager 72 to decide when to reclaim the buffer. The structure of the multi-replication queue is described in Figure 12. In the port vector FIFO 70, the frame index is successfully put into the output queue number 7 of the frame index, and the frame index and the copy number of "> 1" are copied to the multi-copy queue 90. If a specific output queue 74 is full, port direction FI 卩 ϋ 70 0 is not allowed to put a copy of the frame pointer to this output queue; the paper ruler applies the China National Building Standard (CNS > Α4 specifications) (210X297 mm > 30 91472 (please read the precautions before filling out this page) ^ 衣 _ 〇8 0 A7 B7 Printed by the Yuegong Xiaozhou Cooperative of the Central Standards Bureau of the Ministry of Good Yi, 5. Description of the invention (31 ) 7 4; Therefore, it cannot use this event as an example of successfully determining the copy number. Each time the administrator of the puncher 72 takes an output frame indicator and finds that the single copy bit is "0 " (This is a multi-copy action ^ It looks for a frame pointer with a copy number "1" in the multi-output cache, and a frame pointer with a copy number "1" indicates that this is a final input. If Γ is found, after transmitting the contents of each punch, the punch manager 7 2 transmits the frame and recycles the same as a single copy transmission by providing the invalid punch to the free slot 1 0 4 Puncher, if not found, Yuan puncher manager 72 transmits multiple copy frames and Set a copy of the frame index with the copy number "-1" to the multi-copy queue 90. When the host runs out of one, it is arranged in the expansion bus output queue 75 or the management port output queue 77 (via the PCI interface 26). After copying the frame index of the multi-copy frame, the host will write a copy of the frame index with the copy number "-1" into the multi-copy queue via the frame index register. This register is the second A register described in the block diagram of the register 60. Similar to the output queue 74, the multi-copy queue 90 is configured as an input path and an output path. The turn-in path or writing end allows the vector FIFO 70. And the buffer manager puts the frame pointer / copy number to the copy queue 90. The input path, or read end, allows the multiple copy queue 90 to place the frame pointer to the multiple copy cache 96. Provide more information The memory area of the frame pointer / copy number storage capacity is called the multi-copy queue overload area 124. This multi-copy queue overload area is provided by the external memory 36. __ When the frame pointer / copy number is written into a multi-copy Queue 90 mid-hour scale uses Zhongguanjiake {CNS) A4 pit grid (2 10X297 public broadcasting) '~~ 31 91472 -------- W installed ----- ^-order ------ beam (please read the precautions on the back before filling this page) ) 804530 A7 ii7 V. Description of the invention (stops from 32 to TT, full 4 9 terminals, please go to the direct ο 9 columns, storage, multiple entry, write 4, and code 9, consider tl terminal a *, repeat 0 / to the index box. The 2 L, 9SP side writes many books, such as the 12-36 in the C- 2 area, which contains the body memory and the part 94. The outer end of the 94 mark refers to the frame. _ ΙΛ ^ ny ^ 7 ft ft More than 90 times will be listed. Ff will be listed in the system and will be repeated f? Ends will be written in multiple copies. One will be written in the middle of the book. The book will be full. "Du 姶 4 5 2 Open. The next 90 digits will be dequeued and cleared. 4 $ 9 Duplicates have been read and read through. The index frame is duplicated. You have to use the code to duplicate the 12 columns in the book. Duplicate system copy is read from the end of the 9 columns. Duplicate system copy is moved to (阽 Please read the precautions before filling in this book) '' 'Package. The area contains the list of duplicate system copies. Rt- To transfer fast πυ 歹 伫Replicate multiple 1 to provide but ο 9 columns. Replicate multiple types. Take the quick list. Replicate multiple code numbers. / Index frame scan for M area search. Replicate multiple. 2 7 members. Manager Chongyuan C, the material supplier loses 6 5 2 Holds the number system to copy the data according to the root frame, reads the system, reads multiple re-entries, or orders the frame, which is handled by this or 6 9 If the result is as follows, you will lose 74 and pass through the process. 5 2 lose 7 by the staff 2 I 7 the manager manages the buffer buffer.扃 Baygong Consumer Cooperative Co., Ltd. printed and gathered the information management device "„ „1 纟 纟)), how many of the frame number signal system is duplicated? 1 See the indication, 6 Table 9 ί column '伫„ ο system “ Scanning 72 demobilizers for the multi-descriptive position "1 management" management system for the device B 0 镄 鑀. After returning, Ϊ mark to find the fruit box. If the next box is the last one, this table and the input and input of materials Le moved between the transmission and transmission in the staff manager slowly to find I, "No code, no fruit, such as A red square grooves opposing discharge between m after space. After the completion of the process, the unit is completed. 伫 Yuan loses the system. 谡 The reversion is returned to the standard frame. The paper size is applicable to the Chinese National Standard (CNS) Λ4 specification (2) 〇 × 297 mm. 32 9 1472 <? s 〇8Q a7 B7 Printed by M-Ministry Central Bureau of Standards, Industrial Cooperatives, Consumer Cooperatives V. Description of Invention (3 3) 1 1 _The puncher tube member 72 uses the frame index / copy number to perform the task 1 i Multi-copy ff column 9 0 t Put the signal indicator / copy number into the multi-copy cache y 6 1 1 j. Or process the frame indicator / copy number 〇 The m action and frame transmission please read M. Read The side note is no longer relevant. K "'. For example, the two examples are related to the 1) copy number of Yuanyuan administrator_take frame indicator Ί, see if the copy number is" 1 "or " -1 Multi-copy queue 9 0 make take-〇- 涸 copy number 1 1 I 1 1 ** > 1 " frame indicator 〇 Assume that at this time multi-copy cache 96 has storage space 1 installed |, it will __ A new input data is written. If the multi-copy cache 96 does not have %% of this η storage space, the buffer management 72 must clear 1 1 1 of the multi-copy cache 96. The method of clearing space is by fetching multiple copies. Older frame index / copy number in cache 96 1 1 | Use the 1 copy number in this multi-copy cache 96 to update frame index buffer in external memory njn body 36 to order I m, m ι · > · > \ Then delete the input data in the multi-copy cache 96 ♦ I 1 So you can clear a storage t / tr between 0 and a storage space 1 1 1 New The frame index / copy number can be used to write the multi-copy cache 96. 1 2) If the buffer management R 7 2 is multi-copy queue 9 0 capture--a copy number 丨 1 ^ _ | " m index 〇 The M buffer manager 72 searches for multiple m caches 1 1 I 96 to see if there is a frame index copy number a > = 1. Μ 1 ί The next two cases t depend on the yuan Whether the punch manager 72 finds a matching frame index 0 in the multi-copy cache 1 1 take 96. ί a) If m the punch manager 72 finds a matching frame index, 1 * assuming the input of the multi-copy cache 96 The data copy number is a 1 ”, 1 I the buffer tube il member 7 2 will delete the input data t of the multi-copy cache and} 1 I will put the frame pointer into the gate closing queue 98 〇 If more than m Cache 96 1 1 This paper size is applicable to China National Standards (CNS) Λ4 current (210X297 mm) 33 91472 <? ο Θ Ο Α7 Β7 The Ministry of Economic Affairs ’s Standards Bureau lost. 1 Consumption Cooperation Du Yin ¾ 5. Description of the Invention (3 4) 1 I ’s input data copy number is 4 > 1 ”Yuan Chong Administrator 72 Will 1 I reduce the copy number-~ 〇ί 1 b)) Binder management 72 did not find a matching frame index, ίV Please first • I 1 1 This means that the matching frame refers to m has been moved to external memory Read and read the frame 1 of the back of the frame m on the body 36. □ m The buffer management table _ 〇 The buffer management 72 and I 1 vi | You must go to the buffer header and grab this copy number. Matter 1 I The value of this copy number is 1 ", then this frame is no longer needed, fill in 1 again. The punch manager 72 puts this frame index into the recycle queue 98 and writes this page 1 r > if this copy number Value (in memory (Middle) is a > 1 t! I m The punch manager 72 will copy this frame indicator / copy number 1 1 I into the multi-copy cache 9 6 9 and reduce the copy number by 10 if 1 is ordered The multi-copy cache 96 is full. The scale manager 72 moves—1 older frame indicator / copy number to external memory 3 6, and then 1 1 clears out. — Storage space 0 1 1 recycle 伫Column 98 stores frame pointers that point to unneeded link strings. 1 0 Responder manager 7 2 When it serves multiple replication caches 1 It finds that the 1 1 copy number of the indicator is U Γ, > It will write--値 message indicator to the recycle queue 1 1 I 〇 (that is, the last transmission of the frame has been successfully completed) 0 y port vector 1 1 in the case of FIFOS T column ► will also write the message Frame index to re-queue queue 98 1 1 : 1) Although the port direction of the frame index is empty, or 2) because all port vectors' | the transfer queue is full i As a result, the frame indicator cannot be arranged. Finally, when the main m 41 finishes using a single copy frame arranged by the expansion bus port output queue 77 or the management port output 1 column 75, the host will write a frame indicator to recycle 1 Queue 9 8 ί Use frame pointer 搮 register) 0 1 1 This paper size applies to the ten national standard soap (CNS) A4 specifications (2! 0X297 mm) 34 9 1472 、 3 0 8 Λ7 H7 Central Ministry of Economic Affairs Printed by the Consumer Cooperatives of the Standards Bureau. 5. Description of the invention (3 5) 1 1 When the buffer management 72 processes the input data of the recycle queue, it operates one to two r; The punch to idle 1 slot 1 0 The example cell of the present invention does not describe the internal structure of the recovery queue (1 1). In terms of the mm embodiment of the present invention, the recovery mud column only contains the first read — Reading frame I has a frame indicator (14 bit) structure. The writing end of the recovery line has 64 backs. There are 16 input data for the 1 m input terminal in the i return queue. Note 1 1 Similar output queue 7 4 ♦ Recycling queue 98 is structured as input path and output path items and then 1 〇 input path or called writing end ♦ Allow the administrator of the puncher 72 to place the frame% This installation I Pointing to the recovery queue 98 〇 Output m path or called fetching jLiir terminal t allows Yuan punch 1 I admin 72_ fetching frame pointer and return all relevant Yuan punch to empty reading 1 1 1 punch Slot 10 4. More frame indicator storage capacity is provided by the recycle queue overload area 1 2 2 1 1 * This recycle queue overload area is provided by the external memory 36. 1 When the frame pointers are written in «empty collection queue 98, they are passed from the book end 100 to the terminal 100 to the terminal 102, until the terminal 102 is full. For example, 1 1 has more frame pointers to be written into the book end 100 of the recycle queue 98, and 1 person will be located in the overload area 122 of the external memory 36. -S recycling queue 98 1 I's like end 102 and overload area 122 are full, too many 1 1 | put in the recycling queue 98 frame indicator starts to write on the writing end 1 00 〇1 1 Figure 1 1 -An example describing the internal structure of the empty-sinking slot 104 is a 1-cell example, the space buffer slot 104 is--F IF0, which contains--the address index refers to all of i to the external memory 36 Space Yuan puncher 1 40 〇 When all the frames J ί were received * Buffer Manager 72 captures available 1 I address indicators from buffer slot 10 4 新 Stores new incoming data. Yuan Yuan punch Manager 72 also From the space yuan 1 I slot 1 0 4 allocation address to the host processor 2 8 (when required to do so 1 1 this paper size applies to China National Standard (CNS) A4 specifications (2) 0X297 mm) 35 9 1472 4 0 A7 B7 Issued by the Consumer Cooperatives of the Ministry of Standards of the People's Republic of China at the time of the Invention Statement (36). The machine can write or _ fetch a register in the free buffer register 60 that is located in the direct input or output space (cM rec "in f > ti t / ο tp ii tspace). The address index goes to the free buffer slot 104. The writing end 106 and the read end 108 of the free slot 104 of the present invention each hold 64 input data. The empty slot slot 104 is structured as An input path and an output path. (Similar to output queue 74), the input path or writing terminal 106 allows the punch manager 7 2 or host 2 8 to place the address index into the space slot slot) 4 The output path is also referred to as the capture terminal 108, allowing the Yuan Chong administrator 72 to provide the address indicator to the host 28, or to pull out the address indicator M from the space buffer slot 104 for storage and reception of frame data. More storage space of available address indicators is provided by the free slot flushing overload area 120 of the external memory 36, as described above. When the switching device starts to operate, the space flush slot generates the address indicator by the reading end 108 When the frame comes, the free list in the free buffer slot 104 is captured. If there are not enough punch indicators at the writing end 106 to control the data communication, the overload area 120 is accessed to obtain more punch indicators. When the switching device 12 starts to operate, a specific exemplary embodiment of the present invention · Propose the method provided by the more powerful buffer indicator. When the switching device is connected to the power for the first time, the overload area 120 in the external memory 36 does not need to have the buffer indicator. Instead, the buffer indicator is in Generated during the lotus transfer. When the switching device 12 is connected to the power supply, it can generate a buffer index and can put it in the overload area] 2 0, but at this time there are 16 0 0 0 to 3 2 0 0 0 This paper scale is suitable for China National Standards (CNS), \ shirt see grid (210 × 297 public trend) 36 91472 — — -------- #! (Please read the notes of the vg before filling in 3 (This page) < Printed by the Ministry of Economic Affairs, the Central Bureau of Standards, and the Negative Second Consumer Cooperative. V. Invention Description (3 7) The same type of indicator (1), such as a low-cost switching device, is required to start the machine. The present invention adopts the characteristics of -ijb during the opening period. For example, all the m punches are space, and the identification codes of the punches are also known. Therefore, after opening, these buffers are needed. Temple B. Then use the counter 180 to generate these punches. As described in Figure 10. The idle count generator 18 0 (f r e e 1 i s t C 0 u η t gene r a t 〇 Γ) is connected to the input of the multiplexer 1 8 2.茌 When starting up, because the free list in the free slot 10 4 is empty, the space number list generator 1 80 generates the m index 0. Once the free list reaches its highest count, it will be V. Buffer indicators are no longer generated. When a frame packet is received by the switching device 12, the frame packet will be scattered into a fixed-length buffer. * Generally, the frame size will change. These m-bit buffers have 256 bytes, and the part that stores data has 2 40 bytes. After transmitting the contents of the buffer, the indicator of the yuan puncher is put into the recovery queue 98. If the buffer buffer (buffe Γ chai η) can operate *, it is directly placed in the empty punching slot 1 04 〇 In the exchange device 1 2 During the operation, any address index returned to the empty slot 104 is passed from the writing end 106 to the yielding end 10 8 t. If the fetching end 108 is full, many address indices will be written to the external memory. The 36 overload zone 1 2 0 _ 3 0-once the reading end 10 8 and the overload zone 1 2 2 are full 1 and then the address index of the free slot 10 4 will be filled into the space. The writing end of slot 10 4 1 0 6 0 The 13th figure is the root m. The embodiment of the present invention -1 describes the multi-copy queue 9 6 The internal management block representation of the 6 阚 As in the previous simple discussion, enter the multi-copy cache 9 6 The chronological order of input data is maintained. In the present invention, the paper size is applicable to the China National Standard (CNS) Λ4 specification (2 丨 OX 297 mm) 37 91472 53〇A7 B7 M " V. Description of the invention (3 8) »Such as -Wt. M Descritant U > The chronological order is not maintained by the timestamp but the tangerine times of twelve memories tie t · (P hy 3 1 C a Ϊ 0 rderi η S) m keeps η The multi-copy cache of the present invention 9 6 Ιθ] must be m to avoid using the correctness bit, but instead to use the correctness of decoding-this aspect will be discussed in the following M. 0 refer to Figure 13 » Multi-replication cache 96 is constructed as a set of four channels of 1 set of related (S et —as S 0 ciative) memory. Any input data that enters the multi-replication queue 9 6 * contains frame indicators and Copy the number t as mentioned above. The lowest digit of the frame indicator decides whether to store the frame in the relevant cache 96 of each group. Which-line * In the embodiment of the present invention-1, although the size of the cache can be increased, and its number of lines is not limited, but the cache 96 of the present invention has only 6 4 rows. Each m-related cache 96 Divide all the lines into four large lines> Each large line can be searched in parallel. When the buffer manager 72 stores the input data in the cache 9 6, it will always store the uppermost left bit ( 5 1-39), this is specified by the lowest 6 bits of the frame pointer. When this big line is captured, all input data is shifted to the right by 13 bits? And this — the big line will be written back again. 0 The input data that is actually written into the cache 96 * Including the frame indicator most The left 8-bit address mark m (a dd re SS tag) »and the 5-bit eye copy number associated with this frame indicator. 0 When the input data is extracted from the cache, the t frame refers to the» τιττ mark. The frame label and all the bits indicating the cache line 96 will be rearranged. □ If the cache is full and there are new input data to be written, the most input data in the cache will be cached. Removal. As mentioned above, in the case of the m punch 1 42, the copy number of the dedicated frame indicator will be written on the outside of this paper. The size of the paper conforms to the Chinese National Standard (CNS) Λ4 specification (210X297 mm). 38 91472 53〇8〇A7 R7 V. Description of the invention (3 9) The indicator head of the signal receiver pointed out by the frame indicator that was removed during the employment of the case is 14 2. Therefore, the frame is stored in the external memory 3 S (ie, the buffer 140). This external memory can be used to copy the overload area of the cache 9 6 and store the copy number with M. An advantageous feature of the present invention is that there are no correct bits separated in the relevant caches 96. When the copy number is 0 0 0 0 0, the buffer manager 72 knows that this input data is no longer true, so Remove this input from cache 96. This simplifies the structure of the cache. In addition, another advantage of the cache 9R of the present invention is that it allows performing a fast search. Because the buffer manager 72 only needs to check one line, this line is already determined by the frame indicator that left the multi-copy queue 90. The input data of the four major lines are checked in parallel, further increasing the search speed. Although this cache structure is described just like a four-channel group of related memories, this is only an example, which is used to indicate that an n-channel group of related memories does not depart from the scope of the present invention. From the above description, we can know that the present invention can maintain the chronological order of the input data by placing the physical position of the input data in each row (^ ¾ Central Bureau of Standards Bureau employee consumer cooperatives printed β (read first read back Note for both of them, fill in this page again (new and old input data), in other words, the physical location of this input data indicates the new and old level of this input data. To age an input data can be achieved by rearranging the memory. A specific implementation of the present invention, the M-port-by-port basis, can customize the latency of frames switched by the switching device 12. Referring to the 14th deletion, the port to the most FIFO 70 checks the receiving switch's programmed switch mode (this switching device mode is changeable), and M decides when to put the frame index and related information on the paper scale. Commonly used China National Standards (CNS) Λ4 specification (2 丨 0'〆297 mm) 3 9 9 1 472 4 53 0 8 ϋ Α7 ί £ " Printed by Employee Consumer Cooperative of the Central Sample Bureau of Ministry of Justice (4) Enter the output queue 74 of the appropriate transmission port. The first mode (waiting time quality mode) is the most promising direction. IF 0 7 0 There is no limit on the time to broadcast. The indicator is to the output port 7 4 n The second m type (medium waiting time horizontal type) port direction i FIF 0 7 0 Place the frame indicator on the output queue 74, only after the 64 bits of the frame indicator are received. The third m-type (high latency mode) port direction although the PIF ϋ 70 places the frame indicator on the output In queue 74, after all the frame indicators have been received. Here are some special examples. Must the port be changed to pass your frame indicator to the output queue? 4 hours. 1) Transfer the frame from the 100 M b / s port of the first mode or the second mode to the 100 K b / s port. 2) The frame is transferred to the management port 30. 3) The frame is transferred to the expansion bus port. In 1), the speed difference between the 10 Mb / s port and the 100 Mb / s port makes the transfer mode forcibly change to the third high-latency mode. In 2), all the frames passed to the management port use the third mode frame. In 3), the frame transmitted from the M to the expansion bus port uses the switching wedge of the expansion bus port 26. When a multi-port port vector contains one of these special ports, the longest latency switching mode must be used. For example, a frame is received by a port in the first or second configuration, but its multi-copy port vector contains the management port 3 0, so the third horizontal mode must be used to switch modes. In this case, only after all the frame indicators have been received, a copy of a frame finger is placed in the output bank 7 4 0. The switching device will be discussed in more detail horizontally, and the exchange applied to the input port The device mode can determine the transfer waiting time (exchange device 12 2-how fast the K will be transferred to the message frame after receiving the message), and it can be decided to reduce it to each input (read the precautions before reading (Fill in this page) The paper size applies to the national standard (CNS) A4 specification (210X 297 mm) 40 9 1472 45308; A7 printed by the Consumer Cooperatives of the Central Bureau of Standards of the Ministry of Consumers " B7 V. Description of the invention (4 1) The ability to increase the delay / fragment / error · propagation delay „Furthermore, the second anger type, that is, the middle wait time horizontal type is the default mode for each port. And in the register 6 In 0, one port can be used as the head position to regulate the secondary switching device. In the three switching device modes, the frame received by the internal MAC port F 0 5 2 will be transferred to the external memory buffer as much as possible. 1 4 0. At about the same time, rule checker 4 2 or 5 8, receiving purpose Address and source address-receive port vector, frame indicator and some additional information, and then perform the appropriate check „at a glance. After the check is complete, the rule checker 42 or 58 returns the frame indicator and forwards the port vector to FIF 0 7 0 . The port-oriented FIFO 70 places the frame pointer on the writing terminal 76 of the output queue 74, and provides the port-oriented view M to identify the output port. When the port vector FIFO 70 receives the port vector (frame indicator) and puts the frame indicator in the output queue 74, the switching device mode of the receiving port defines the transfer waiting time of the port. This is a narrative for the following three patterns. Once the frame indicator is passed to the reader 7 8 of the output queue 7 4, the buffer manager 7 2 _ takes the frame indicator and schedules the transmission time. The buffer manager 72 starts moving the frame data by the address indicated by the frame pointer. Once the transmission of the FIF054 of the MAC port is back to the point where it first started (assuming the media can be used to transmit frame data), the transmission of the frame begins. The first mode is designed to provide a transmission mode with the lowest latency. The transmission of the frame and the speed of transmission at the M line rate (1 i ti e-rated speed) are used for transmission. Before confirming that the frame is a block (ie, less than 2 R bits) or before including a CRC error, arrange K for this paper size to apply the Chinese National Standard {CNS} 4 specifications (2) 0: < 297 mm) 41 91472 (Please read the precautions before reading the '" this page) 4 53 0 A7 R7 Printed by the Consumer Cooperatives of the Central Standards Bureau of the Ministry of Economic Affairs of the People's Republic of China 5. Description of the invention (4 2 ) Transmission, so there is no network error _ protection (netw 〇rkerr 〇r P "otect; ί ο n). In the first mode, before the frame of the output port is transmitted, the frame reception may not be completed. If The reception of the frame ends with an immature transmission period (runt), or with an incorrect CRC, the receiving MAC will indicate < ma "k > buffer 1 4 2 of external memory 36, K shows that these are happening now. Fu loses Mac guarantees that if a frame starts to be transmitted, this frame will then end with an immature Fu Input cycle. Or if there is an incorrect CRC, Μ Λ C will generate a bad CHC. The MAC has not yet started the transmission of the frame, but the buffer 142 has indicated that the frame ends at an immature transmission cycle or has an incorrect CRC. The buffer manager will not forward the frame to the output port. The second horizontal mode provides a low-latency frame forwarding horizontal mode, but with a little network error protection. Upon receiving 64-bit or more bits The frame will only be forwarded afterwards. This will allow the switching device 12 to filter out (ie not forward) the frame of the segment. However, this mode cannot completely pass over frames with more than 64 bits and CRC errors. Second horizontal , The frame index of the frame that reaches the B 4-bit critical value is arranged to the appropriate output queue 7 4. The frame that does not reach the 64-bit critical value will be deleted, and its signal indicator will not be put out. Output queue 7 4. If one is greater than or equal to The 5 4-bit frame ends with an incorrect CRC, receiving the MAC to mark the external memory 3 (3's buffer 1 4 2, Μ indicates that this is currently happening. If the transmission starts with a greater than or equal to 64-bit frame. Then it ends with an incorrect C 1? C. Transmission Μ Λ C will end with a bad CKC. If a flf transmission H / \ C module does not start transmitting frames, but --------- Installation -------- Order. (诮 Please read the precautions on the reverse side and fill in the "Smart page") This paper size is applicable to China National Standard (CNS) Λ4 specification (210X297 Mm) 42 9 1472 4 53 Ο

C Α7 Β7 五、發明説明(w Μ 列 Sfff 框管製 訊器複 此衝多 示 緩戎 0 ) 2 C 送 4 DK f 1 C 轉 器的 製 衝確 複 猙正單 於 等 或 於 大 元 位 不 個 98送 列轉 一 庁 必 於收 不 束回而 結到 -丨搮丨 指 框 訊 遨 歸 送 轉 製 複 多 列 佇 出 輸 至 誤 錯 路 網 的 準 水 高 最 有 具 中式 模 個三 是 式三 第 式 送 轉 存 儲 的 間 時 待 等 送 轉 高 '_ 保 Θ Γ Γ ο Γ 3 埠 出 輸 至 HJ f 它 送 轉 要 將 2 1± 置 裝 換 交 在 框 訊 之 送 轉 再 2 1—_ 置 裝 換 交 0 收 接 被 全 完 已 之 的 有 所 絕 隔 (讀先間讀背而之注意事項甩填巧本頁 有 所 及 段 Η 下 式 模 <Ϊ1 種三 第 在 框 訊 框 訊 個1旦 的 確 正 著 有 Μ 元 位 4 6 於 等 或 於(* 端 收 接 於 束 结 地 功 成 歹 庁 出 輸 的 當 適 到 ^ϋ 歹 OU- 地 功 成 會 標 指 榧 訊 結 柢 訊 的 確 正 一八/1- 誤 錯 收 接1 於 束 被 會 將 )/ 等 期 週 的 熟 成 不 列 佇 出_ 人 放 會 不 也 標 指 框 訊 的 們 它 除 刪 ο 7 接 施值元 與 實界位 , 的臨六 定 別個如 決在三ΐ( 的,有 位 列然 ,η 佇 雖言收 出 。例接 輸關施1) 入 無實是 放 量本值 量料Μ界 向資 ,臨 埠 的值些 作 收界這 接臨 , 及的中 F0式目例 FI模數腌 量 的同實 向 擇不範 埠 選有示 埠中本 收例 。 經济部中央標準扃負工消貧合竹社印袈 於 Π 中 其 框 訊 的 有 所 收 接 3 元 位 4 6 收 接 列 ft 出 輸 至 框 訊 送 轉 值 界 臨 據 根 明 發 本 景 向 埠 安 新S 式m 的 則 規 被 埠 該 及 最 料 資 的 收 接 據 根 但根 , 如 定 例 決 , 的 定 送決 轉送 作轉 量作 料素 資因 的他 收 其 接 據 康 Μ ί 根卻 然 例 雖施 C 莨 序 呰 次有 的明 輸 發 傳本 排在 態 型 的 料 資 收 接 撺 員 3一| ί'管器ir 鍰 時 規 送 轉 明 發 本 成完 在 取 快在 本紙张尺度適用中國國家標準(CNS ) A4規格(210>< 297公釐) 91472 4 53 〇8 Α7 Β7 經濟部中央標準局貝工消費含竹杜印5;! 五、發明説明(4 4) 記憶體1 6]中,保持著一個表1 6 0。該表使訊框指搮與接收 埠發生關連。每次埠向量F i F 0 7 0由規則檢查器4 2或5 8接 收一俩新的埠向最和訊框指標,它會產生一涸關聯來確定 該輸人埠是否已完全接收到一框訊,如果沒有,則確定有 多少訊榧被接收到。埠向t F I F 0 7 0並沒有由檢查器4 2或 5 8接收到仟何有關於接收坶的識別賣料,埠向量F I F 0唯 一接收到有關於接收埠的識別資料是訊框指標。 不管是在訊柢尚在接收,位址表歸邀接收埠的情況, 或是因訊框已被完全接收,導致位址表1 6 0找不到訊框指 標的情況,埠向量FI F 0 7 0會以訊框指標來査詢位址表1 6 0 。一旦訊框完全被接收,其訊框指標會被移出位址表160 。此種情況表示碰到了第三種模式(訊框完全接收)。因而 訊框指標立即掉人輸出佇列中。 如果位址表1 6 0歸遒到接收埠,埠向景F I F 0 7 0將訊框 指標及相關資訊放入保持區162,並開始由接收埠監視兩 個信號。此兩個信號搮示(flag)三個事件中的一個。第一 個事件標示當埠接收到η個位元。在此一時間點,如果這 一個埠茌第一個模式,則埠向量F I F 0 7 0賴著將它送到適當 輸出佇列74而開始處理此訊框。如果不是茌第一個模式, 埠向最F I F ϋ 7 0會等待下去,直到它接到一個信號標示著 遇到Τ第二種事件。如果該埠是在第二種模式,埠向1 l· I F 0 7 0會將訊框指標由保持區1 6 2釋放到適當的輸出疗列 74。最後,如果該坶是在第三種樓式,埠向量PIF0 70會 等待,直到他接到一個旗標表示訊框已完全被接收。每一 IN氏张尺度適用中國國家標準(CNS ) Λ4規格(2丨0X297公釐) 44 91472 {讳先閱讀背面之注意事項再填寫本頁) 4 53080 Λ7 Β7 五、發明説明(45) 個接收堉(參考第14圖數字ΐ (5 4者)會維捋這個旗標,並提 供這一個資訊給埠向虽FI F 0 7 0。此旗標提供資訊給埠向 量F I F 0 7 0以確定該埠關聯於哪一個訊框指標。而埠向量 F 1 F 0 7 0會保持該埠是在何種_式的資訊。總之,當接收 一個訊框指標的時候,埠向最F I F () 7 0資詢缓衝器管理員 7 2的位址表1 (5 0以決定是哪一個接收埠,確定此接收埠的 m式,由該接收埠監視一些旗標,根撺該横式和這_旗標 去釋放訊框指標。 雖然本發明已詳細解說及揋示,應清楚地知道這些範 例俺用來說明本發明並非用來限制本發明。本發明之精神 及範赌,將Μ所附之申請專利範園之條款來限制。 (誚先間讀背而之注意事項再填寫本哀) 經满部中央標準局员工消費合竹社印袈 9 1472 本紙張尺度適用中國囤家標準(CNS } Α4規格(210X 297公釐)C Α7 Β7 V. Description of the invention (w MW column Sfff frame control signal repeats this to show more time 0) 2 C send 4 DK f 1 C converter to make sure that the order is equal to or equal to Yuan If the number is not 98, it will be transferred if it is not received.-丨 搮 丨 Finger frame information 遨 Return and transfer to multiple columns 伫 The highest level of quasi-water level that is lost to the wrong network has the most Chinese model. The third is the type of the third type of transfer and storage. Waiting for the transfer to be high for a while'_ Guarantee Θ Γ Γ ο Γ 3 Port output to HJ f It sends and transfers 2 1 ± installation and exchange in the frame of the message Then 2 1—_ install, change, and hand over. 0 The reception is completely separated. (Notes for reading before reading the back. Please fill in this page.) The following formula < Ϊ1 kind of three In the frame, a frame of 1 frame is indeed marked with a M bit of 4 6 equal to or equal to (* end received at the end of the ground power success 歹 庁 输 歹 歹 OU- Earth power success monogram Indicating that the final message is indeed 18/1-mistakenly received 1 Yu Shu will be) / waiting period The maturity is not listed _ Ren Fanghui also refers to those who frame the message. It deletes ο 7 the value of the element and the real boundary. η 伫 Although it is closed. For example, if you enter the policy, you will be able to measure the value of the M-valued data, and the value of the port will be used as the closing value. The example of picking the same amount of reality is to show the example of the central bank. The central standard of the Ministry of Economic Affairs, the Ministry of Economic Affairs and the Poverty Alleviation Co., Ltd., printed in the box, has received 3 yuan, 4 6 yuan. The output of ft to the frame and the transfer of value to the border is based on the fact that the present situation to the port of the new S-m is ruled by the port and the most valuable receipt, but as the rule of law, He decided to transfer the data as the source of the material, and he received the receipts. However, the example is based on the C code, and there are some clear-cut and circulated materials. | ί'tube ir 锾 规 转 送 送 送 送 转 明 明 明 明 明 明 明 本 本 成 完 取 取 取 取 取 取 快 快 适用 适用 适用 Applicable to this paper size National Standard (CNS) A4 Specification (210 > < 297 mm) 91472 4 53 〇8 Α7 Β7 Shellfish Consumption by the Central Standards Bureau of the Ministry of Economy Contains Bamboo Du Yin 5 ;! 5. Description of the Invention (4 4) Memory 1 6], maintaining a table 1 6 0. The table associates the frame pointer with the receiving port. Each time the port vector F i F 0 7 0 is received by the rule checker 4 2 or 5 8 a new two-way frame index, it will generate a correlation to determine whether the input port has completely received a frame If not, determine how many messages have been received. The port direction t F I F 0 7 0 is not received by the inspector 4 2 or 5 8, so there is no identification information about the receiving party. The port vector F I F 0 uniquely receives the identification information about the receiving port is a frame indicator. No matter whether the address table is still receiving, the address table is invited to receive the port, or the frame table has been completely received, resulting in the address table 1 6 0 can not find the frame index, the port vector FI F 0 70 will use the frame index to query the address table 160. Once the frame is completely received, its frame index will be removed from the address table 160. This situation indicates that a third mode is encountered (the frame is fully received). Therefore, the frame indicator immediately dropped into the output queue. If the address table 160 is assigned to the receiving port, the port view F I F 0 7 0 puts the frame indicator and related information into the holding area 162, and starts to monitor two signals by the receiving port. These two signals flag one of three events. The first event indicates that the port received n bits. At this point in time, if this port is the first mode, the port vector F I F 0 7 0 relies on sending it to the appropriate output queue 74 to begin processing this frame. If it is not the first mode, port direction F I F ϋ 7 0 will wait until it receives a signal indicating that the second event is encountered. If the port is in the second mode, the port direction 1 l · I F 0 7 0 will release the frame pointer from the holding area 16 2 to the appropriate output therapy row 74. Finally, if the frame is in the third building, the port vector PIF0 70 will wait until he receives a flag indicating that the frame has been completely received. Each IN scale applies the Chinese National Standard (CNS) Λ4 specification (2 丨 0X297 mm) 44 91472 {Please read the precautions on the back before filling this page) 4 53080 Λ7 Β7 5. Description of the invention (45) Receive堉 (Refer to the number ΐ (5 4 of 14) in Figure 14 will maintain this flag and provide this information to the port direction FI F 0 7 0. This flag provides information to the port vector FIF 0 7 0 to determine the Which frame indicator is associated with the port. And the port vector F 1 F 0 7 0 will keep the information of what type the port is. In short, when receiving a frame indicator, the port direction is up to FIF () 7 0 Consult the buffer manager 7 2's address table 1 (50 to determine which receiving port, determine the m-type of the receiving port, and the receiving port monitors some flags, based on the horizontal and this_ Flag to release the frame indicator. Although the present invention has been explained and illustrated in detail, these examples should be clearly understood to illustrate that the present invention is not intended to limit the present invention. The spirit and scope of the present invention are attached to M The terms and conditions of the patent application park are restricted. (Please read the precautions first and then fill out this guide) Department of the Central Bureau of Standards staff full co-consumption society bamboo paper printed Buddhist monk's robe 91472 Ben scale applicable standard home store China (CNS} Α4 specifications (210X 297 mm)

Claims (1)

〇 8 ο 3 54 ABCD 六、申請專利範圍 法 方 的 輸 傳 増 画 複 -:u 經 框 料 賣 的 製 複 制 控 t—··' 榑 驟 的 '卜 以 拮 包 0 IV 515 的 製 複 輸 傳 S- 此二 哪 由 將 定 指 製 複 - 其 輸 傳 來 用 可 不 掉 些 哪 有 中 埠 的 定 及 指以 定 確訊 框 的 定 指 β 框 訊 製 複 其 輸 傳 來 用 可 到 框 訊 製 複 供 提 複 將 中 坶 定 指 的 框 訊Ρ 複 輸 傳 來 用 可 不 在 中 埠 棄 放 框 訊 之 生0 址 位 的 目 的 中 框 據 根 括 包 驟 第 步 圃 的 範框 專的 請 製 申 複 如輸 2 法 方 的 項 傳 埠 些 哪 由 將 定 指 中 其 埠 定 指 些 這 別 0 會 將 虽 向 埠 而量 向 埠 個 - 相 個1 有 埠 個1 每 中 其 法 方 的 項 2 第 圍 範 利 專 請 甲 如 3 使 一 輸 鱼 勺 , .Β Ο 標 更埠方 指 法關地 框 方相的 訊 該人去 列 .載出 排 址標輸 剌 位指傳 規體 框此 被 憶訊由 列 記 的將 佇 的框框 出 框訊訊 輸 訊的為 該 存送即 , 儲傳堉 列 到要些 佇指 將該 出標 , , 輸 指括列 的框 包佇 闞 訊步出 當 括 包 步 I 進 更 法 方 該 法 方 的 項 3 第 圍 範 利 專 請 & 如 -請先Μ讀背面之注意事項再填寫本頁) 經濟部中央標準局—工消f合作社印製 候 時 的 開 離 列 佇 出 輸 的 埠 關 相 由 標 指 框 訊 的0 ΪΪ5 埠 的 框 訊 收 固 體 憶 記 中 埠 的 定 指 定 確 中 其 法 方 的 項 4 第 圍 範 禾 專 請 Φ 如 5 檢 括 包 驟 步 之 框 的 製 複1 其 輸 傳 來 用 可 不 埠 些 哪 有 出 輸 ηκ 相 的 埠 個 - 當 中 其 量 容 用 利 可 的 HJ 歹 佇 出 輸 査 輸 傳 來 用 不 定 確 被 將 埠 該 時量 容 用 利 可 有 沒 已 列 佇 框 ΤίΗ 的 複 - 其 本紙張尺度適用十國國家標準(CNS ) Λ4規格(2!0X297公釐) 46 91472 A8 B8 C8 D8 經濟部中央標準局員工消費合作社印製 六、申請專利範圍 1 1 6 . 如 請 專 利 m 圍 第 5項之方法, 其中丟棄將要傅輸的複 1 1 製 訊 框 的 步 锞 包 括 丟 m 將 要 載 入 無 可 利 用 容 量 的 指 定 1 I 埠 的 相 關 輸 出 佇 列 的 訊 m 指 標 之 步 驟 〇 請 1 先 1 7 . 如 Φ m 專 利 m 圍 第 G 項 的 方 法 + 更 m — 步 钽 括 確 定 被 闼 1 背 ί 丟 棄 的 訊 框 指 m 數 冃 〇 之 ί 注 1 8 . — 揷 多 埠 網 路 交 換 裝 貿 * 該 多 埠 網 路 交 換 裝 贊 傳 輸 單 意 事 1 項 1 一 個 賁 料 訊 框 之 複 數 個 副 本 1 該 網 路 交 換 裝 置 包 括 再 1 複 數 俩 埠 將 要 傳 輸 的 資 料 訊 框 由 該 處 傳 輸 I fi 本 頁 裝 1 一 個 埠 向 最 產 生 器 用 Μ 產 生 璋 向 量 此 埠 1 1 向 最 會 指 定 將 要 傅 輸 的 訊 框 將 由 哪 埠 傳 輸 , 1 I —- 個 傅 輸 的 規 劃 用 Μ 決 定 這 些 被 埠 向 量 指 定 傳 1 i 訂 輸 單 — 個 複 製 訊 框 之 埠 的 利 用 率 « 及 由 指 定 埠 中 可 用 1 傳 輸 複 製 訊 框 之 埠 的 傳 輸 複 數 個 複 製 訊 框 t 丢 掉 複 數 1 1 個 複 製 訊 框 的 動 作 » 則 由 被 指 定 傳 輸 複 製 訊 框 的 埠 中 1 1 不 能 用 來 於 傳 輸 複 製 訊 榧 的 埠 來 做 Λ 1 線 9 . 如 請 專 利 範 園 第 8 項 的 網 路 交 換 裝 置 > 其 中 的 傳 輸 1 | 規 制 包 括 複 m 個 輸 出 佇 列 t 其 中 有 個 特 殊 的 輸 出 佇 1 I 列 相 關 連 於 一 個 特 殊 的 埠 0 1 1 10 ,如 請 專 利 範 園 第 9項的網路交換裝置, 其中的傳輸規 1 ! 剌 包 括 個 埠 向 最 個 F IF0 被 規 劃 為 接 收 埠 向 最 產 生 器 1 1 t 及 確 定 踉 指 定 埠 有 關 的 每 一 個 輸 出 佇 列 的 可 利 用 容 1 1 最 Γ> 1 I 11 .如 請 專 利 範 圍 第 1 0項 的 網 路 交 換 裝 置 t 其 中 埠 向 量 i I FIFO?! 進 一 步 被 規 剌 為 * 當 某 一 輸 出 fr 列 有 可 利 用 容 I i 本紙張尺度適用中國國家標準(CNS ) Λ4規格{ 210X297公釐) 47 91472 3 5 ο 8 ABCD 六、申請專利範圍 量時,對每一將要傳輪的複製訊框,載入訊框指標到 跟指定埠有關的鑰出佇列;訊框指標指到記憶體中儲 存訊框的位址,输出庁列排列訊框指檷,其中每一個 埠被規剌為存取該被訊框指摞指到的記憶體位址,並 根據離開跟該埠有關的輪出佇列的訊框指標,傳輸在 該位址的訊框。 12. 如申請專利範圍第11項的網路交換裝置,其中埠向量 FIFO更進一步被規剌為如果输出佇列沒有可利用容量 時,丢棄跟指定埠有關的輪出佇列中的訊框指標。 13. 如申請專利範圍第12項的網路交換裝置,其中埤向量 FIFO被進一步規刺為可確定一涠特定的訊框中,被載 入輸出佇列中訊框指標的數目,及確定一個特定的訊 框中,所放棄的訊框指標的數目。 ---------t------IT------it ‘.(請先闳讀背面之注意事項再_ X'本頁) 經濟部中央標準局員工消費合作社印製 本纸張尺度適用中國國家標準(CNS) A4洗格(2丨0X297公釐) 48 91472〇 8 ο 3 54 ABCD VI. Application for Patent Scope: France's Loss and Transfer of Reproduction-: u Copy Control of Selling through Frame Materials t— · '' Steps' to Encourage Reproduction of 0 IV 515 Biography S- The two which can be used to control the dictation-its input can be used without the use of which can not be fixed, and the definite sign of the definite frame β frame can be used to control its input and transmission The message system provides for the retransmission of the frame message P, which is the fixed reference point. The frame frame P is included in the frame of the first step. Please make a reply if you lose 2 items in the French method, which will be set in the port, and will be set in the port. 0 will be directed to the port although it is directed to the port-one, one, and one each. Item 2 Fan Li specially asked Jia Ru 3 to make a fish spoon, .B 〇 The bidder ’s fingering the frame and the frame of the box, the person should be listed. The address of the input box is set to pass the body frame. This is remembered by the listed general The framed information is the storage and transmission, that is, the storage and transmission of the information to the target and the bidding, and the input of the framed information includes the step of the information, and the method of entry. Item No. 3 of the legal party: Fan Li Special & If-Please read the notes on the back before filling out this page) Central Standards Bureau of the Ministry of Economic Affairs-Industrial Consumers Cooperative Cooperatives The phase of the port is identified by the frame of 0 ΪΪ5. The frame of the port is received in solid memory. The designation of the port is confirmed in the legal method of the port. 4 Fan Fanhe specially invited Φ as 5 to include the box of the step Remedy 1 The input and output can be used. What can be used to export the ηκ phase?-Among them, the volume can be measured by Leica's HJ. The output can be used to check the transmission. A copy of the box is not available-the paper size is applicable to the ten national standards (CNS) Λ4 specification (2! 0X297 mm) 46 91472 A8 B8 C8 D8 Printed by the Consumer Cooperatives of the Central Standards Bureau of the Ministry of Economic Affairs 1. The scope of application for patents 1 1 6. If you ask for the method of item 5 around m, where the steps of discarding the complex 1 1 frame that will be lost are included, including the loss of m to be loaded into the designated 1 I port with no available capacity. The steps of the relevant output index of the m index are 〇 please 1 first 17. For example, the method of Φ m patent m around the G item + more m — Step tantalum to determine the frame discarded by 闼 1 back refers to the number of m 冃. Zhi Note 1 8. — 揷 Multi-port network exchange equipment trade * The multi-port network exchange equipment transmission transmission single note 1 item 1 a plurality of copies of an information frame 1 the network exchange device includes another 1 plural The data frame to be transmitted by the two ports is transmitted from there I fi This page is installed 1 A port is generated by the generator with a vector M This port 1 1 is assigned to the most Which port will be transmitted by the frame, 1 I —- The plan of the four inputs is used to determine those designated by the port vector. 1 i Order form — the utilization of the port of a duplicate frame. The transmission of the copy frame port transfers multiple copies of the frame t to lose the plural 1 1 copy frame action »then the port designated to transfer the copy frame 1 1 cannot be used to transfer the copy frame to the port Λ 1 line 9. For example, please request the network switching device of item 8 of the patent > transmission 1 | The regulation includes multiple m output queues t of which there is a special output 伫 1 I is related to a special port 0 1 1 10, for example, the network switching device of item 9 of the patent fan park, where the transmission rule 1! 剌 includes the port to the most F IF0 is planned to receive the port to the generator 1 1 t and confirm the designation Port-related The available capacity of each output queue is 1 1 to Γ > 1 I 11. For example, the network switching device of the patent scope item 10, where the port vector i I FIFO ?! is further defined as * when a certain output fr Available paper sizes are listed. The paper size is applicable to Chinese National Standard (CNS) Λ4 specification {210X297 mm) 47 91472 3 5 ο 8 ABCD. 6. When applying for the scope of patents, copy the frame for each transmission. Load the frame pointer to the key queue associated with the specified port; the frame pointer refers to the address of the storage frame in the memory, and the output queue array frame pointer, where each port is defined as access The framed frame refers to the memory address pointed to, and the frame at the address is transmitted according to the frame index leaving the rotation queue related to the port. 12. For the network switching device under the scope of application for patent No. 11, the port vector FIFO is further regulated to discard the frame in the rotation queue related to the designated port if the output queue has no available capacity. index. 13. For example, the network switching device under the scope of application for patent No. 12, in which the vector FIFO is further specified to determine a specific frame, the number of frame indicators loaded into the output queue, and determining a The number of dropped frame metrics for a particular frame. --------- t ------ IT ------ it '. (Please read the precautions on the back before _ X' this page) Staff Consumer Cooperatives, Central Bureau of Standards, Ministry of Economic Affairs Printed paper sizes are in accordance with Chinese National Standard (CNS) A4 (2 丨 0X297mm) 48 91472
TW087101988A 1997-02-14 1998-02-13 Method and apparatus for selectively discarding packet for blocked output queues in the network switch TW453080B (en)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
US3802597P 1997-02-14 1997-02-14
US08/992,428 US5953335A (en) 1997-02-14 1997-12-18 Method and apparatus for selectively discarding packets for blocked output queues in the network switch

Publications (1)

Publication Number Publication Date
TW453080B true TW453080B (en) 2001-09-01

Family

ID=26714735

Family Applications (1)

Application Number Title Priority Date Filing Date
TW087101988A TW453080B (en) 1997-02-14 1998-02-13 Method and apparatus for selectively discarding packet for blocked output queues in the network switch

Country Status (6)

Country Link
US (1) US5953335A (en)
EP (1) EP0960503B1 (en)
JP (1) JP4603102B2 (en)
DE (1) DE69803364T2 (en)
TW (1) TW453080B (en)
WO (1) WO1998036529A1 (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US7899924B2 (en) 2002-04-19 2011-03-01 Oesterreicher Richard T Flexible streaming hardware

Families Citing this family (113)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6061351A (en) * 1997-02-14 2000-05-09 Advanced Micro Devices, Inc. Multicopy queue structure with searchable cache area
US6175902B1 (en) * 1997-12-18 2001-01-16 Advanced Micro Devices, Inc. Method and apparatus for maintaining a time order by physical ordering in a memory
US6185630B1 (en) * 1997-02-14 2001-02-06 Advanced Micro Devices, Inc. Device initializing system with programmable array logic configured to cause non-volatile memory to output address and data information to the device in a prescribed sequence
US6064805A (en) * 1997-07-02 2000-05-16 Unisys Corporation Method, system, and computer program product for intraconnect data communication using buffer pools and buffer pool management
US6345296B1 (en) 1997-07-02 2002-02-05 Unisys Corporation Method system and computer program product for providing pull model data communication
JPH11127188A (en) * 1997-10-20 1999-05-11 Fujitsu Ltd Information transmission system and method in store and forward exchange type electronic conference system and medium stored with information transfer program
US6178414B1 (en) * 1997-12-16 2001-01-23 Nortel Networks Limited Method and apparatus for updating and searching an ordered list of values stored within a memory resource
US6560227B1 (en) * 1998-02-23 2003-05-06 International Business Machines Corporation LAN frame copy decision for LAN switches
US6246690B1 (en) * 1998-03-19 2001-06-12 3Com Corporation Method and device for controlling data flow in a computer data network
US6092108A (en) * 1998-03-19 2000-07-18 Diplacido; Bruno Dynamic threshold packet filtering of application processor frames
US6473424B1 (en) * 1998-12-02 2002-10-29 Cisco Technology, Inc. Port aggregation load balancing
US6934293B1 (en) 1998-12-02 2005-08-23 Cisco Technology, Inc. Port aggregation load balancing
US6810042B1 (en) * 1999-01-04 2004-10-26 Cisco Technology, Inc. Method of queuing traffic to many destinations covering a wide range of transmission speeds while minimizing buffer size
US6487199B1 (en) * 1999-04-07 2002-11-26 Advanced Micro Devices, Inc. Method and apparatus for maintaining randomly accessible copy number information on a network switch
US7369550B1 (en) 1999-04-22 2008-05-06 Advanced Micro Devices Method and apparatus for locking a table in a network switch
US6529503B1 (en) * 1999-05-21 2003-03-04 Advanced Micro Devices, Inc. Apparatus and method for storing header information in a network switch
US6401147B1 (en) * 1999-05-24 2002-06-04 Advanced Micro Devices, Inc. Split-queue architecture with a first queue area and a second queue area and queue overflow area having a trickle mode and an overflow mode based on prescribed threshold values
US6751191B1 (en) 1999-06-29 2004-06-15 Cisco Technology, Inc. Load sharing and redundancy scheme
US7065779B1 (en) * 1999-10-13 2006-06-20 Cisco Technology, Inc. Technique for synchronizing multiple access controllers at the head end of an access network
GB2355374B (en) * 1999-10-15 2001-10-31 3Com Corp Network device including selective discard of packets
US6700897B1 (en) 1999-10-29 2004-03-02 Advanced Micro Devices, Inc. Apparatus and method for identifying data packet types in real time on a network switch port
US6798788B1 (en) 1999-11-24 2004-09-28 Advanced Micro Devices, Inc. Arrangement determining policies for layer 3 frame fragments in a network switch
US6697330B1 (en) * 1999-11-26 2004-02-24 Hewlett-Packard Development Company L.P. Method and system for output flow control in network multiplexers
US6697380B1 (en) 1999-12-07 2004-02-24 Advanced Micro Devices, Inc. Multiple key lookup arrangement for a shared switching logic address table in a network switch
US6728246B1 (en) 1999-12-07 2004-04-27 Advanced Micro Devices, Inc. Arrangement for reducing layer 3 header data supplied to switching logic on a network switch
US6950434B1 (en) 1999-12-07 2005-09-27 Advanced Micro Devices, Inc. Arrangement for searching packet policies using multi-key hash searches in a network switch
US7103035B1 (en) 2000-01-14 2006-09-05 Advanced Micro Devices, Inc. Arrangement for searching network addresses in a network switch using multiple tables based on subnet identifier
US7058007B1 (en) 2000-01-18 2006-06-06 Cisco Technology, Inc. Method for a cable modem to rapidly switch to a backup CMTS
US6839829B1 (en) 2000-01-18 2005-01-04 Cisco Technology, Inc. Routing protocol based redundancy design for shared-access networks
US6574240B1 (en) 2000-01-19 2003-06-03 Advanced Micro Devices, Inc. Apparatus and method for implementing distributed layer 3 learning in a network switch
US6745246B1 (en) * 2000-01-28 2004-06-01 Advanced Micro Devices, Inc. Apparatus and method in a network switch for modifying a bandwidth request between a requestor and a router
US6934260B1 (en) 2000-02-01 2005-08-23 Advanced Micro Devices, Inc. Arrangement for controlling learning of layer 3 network addresses in a network switch
US6711161B1 (en) 2000-02-24 2004-03-23 Advanced Micro Devices, Inc. Arrangement for providing linearly scaleable address forwarding tables within multiple network switch modules
US6760341B1 (en) 2000-02-24 2004-07-06 Advanced Micro Devices, Inc. Segmention of buffer memories for shared frame data storage among multiple network switch modules
US7002955B1 (en) 2000-03-06 2006-02-21 Advanced Micro Devices, Inc. Selective address table aging in a network switch based on application state determined from a received data packet
US6674769B1 (en) 2000-03-07 2004-01-06 Advanced Micro Devices, Inc. Simultaneous searching of layer 3 policy filter and policy cache in a network switch port
US6732209B1 (en) * 2000-03-28 2004-05-04 Juniper Networks, Inc. Data rate division among a plurality of input queues
US6862265B1 (en) 2000-04-13 2005-03-01 Advanced Micro Devices, Inc. Weighted fair queuing approximation in a network switch using weighted round robin and token bucket filter
US6807179B1 (en) 2000-04-18 2004-10-19 Advanced Micro Devices, Inc. Trunking arrangement in a network switch
US7079537B1 (en) 2000-04-25 2006-07-18 Advanced Micro Devices, Inc. Layer 3 switching logic architecture in an integrated network switch
US6714542B1 (en) 2000-05-01 2004-03-30 Advanced Micro Devices, Inc. Apparatus and method for storing min terms in a central min term memory for efficient sharing by a plurality of network switch ports
US6571291B1 (en) 2000-05-01 2003-05-27 Advanced Micro Devices, Inc. Apparatus and method for validating and updating an IP checksum in a network switching system
US6693908B1 (en) 2000-05-01 2004-02-17 Advanced Micro Devices, Inc. Apparatus and method for efficient evaluation of equations which generate frame tags in a network switch
US6807183B1 (en) 2000-05-09 2004-10-19 Advanced Micro Devices, Inc. Arrangement for reading a prescribed location of a FIFO buffer in a network switch port
US6742044B1 (en) 2000-05-10 2004-05-25 Cisco Technology, Inc. Distributed network traffic load balancing technique implemented without gateway router
US6789125B1 (en) 2000-05-10 2004-09-07 Cisco Technology, Inc. Distributed network traffic load balancing technique implemented without gateway router
US7502366B1 (en) 2000-05-23 2009-03-10 Advanced Micro Devices, Inc. Arrangement in a network switch for prioritizing data frames based on user-defined frame attributes
US6678272B1 (en) 2000-05-24 2004-01-13 Advanced Micro Devices, Inc. Apparatus and method using a register scheme for efficient evaluation of equations in a network switch
US6757742B1 (en) 2000-05-25 2004-06-29 Advanced Micro Devices, Inc. Computer-based system for validating hash-based table lookup schemes in a network switch
US7395348B1 (en) * 2000-06-05 2008-07-01 Cisco Technology, Inc. Network cache-based content routing
US6981054B1 (en) 2000-06-06 2005-12-27 Advanced Micro Devices, Inc. Flow control arrangement in a network switch based on priority traffic
US6925085B1 (en) 2000-06-07 2005-08-02 Advanced Micro Devices, Inc. Packet classification using hash key signatures generated from interrupted hash function
US6718379B1 (en) 2000-06-09 2004-04-06 Advanced Micro Devices, Inc. System and method for network management of local area networks having non-blocking network switches configured for switching data packets between subnetworks based on management policies
US6678276B1 (en) 2000-06-15 2004-01-13 Advanced Micro Devices, Inc. Apparatus and method for specifying successive byte locations for evaluating data packets in a network switch port
US6741594B1 (en) 2000-06-15 2004-05-25 Advanced Micro Devices, Inc. Arrangement for identifying data packet types from multiple protocol formats on a network switch port
US6711165B1 (en) 2000-06-15 2004-03-23 Advanced Micro Devices, Inc. Apparatus and method for storing min terms in network switch port memory for access and compactness
US7072979B1 (en) 2000-06-28 2006-07-04 Cisco Technology, Inc. Wide area load balancing of web traffic
US6981056B1 (en) 2000-06-28 2005-12-27 Cisco Technology, Inc. Wide area load balancing of web traffic
US7062571B1 (en) 2000-06-30 2006-06-13 Cisco Technology, Inc. Efficient IP load-balancing traffic distribution using ternary CAMs
US7069324B1 (en) 2000-06-30 2006-06-27 Cisco Technology, Inc. Methods and apparatus slow-starting a web cache system
US6807176B1 (en) * 2000-07-13 2004-10-19 Advanced Micro Devices, Inc. Arrangement for switching data packets in a network switch based on subnet identifier
US6714556B1 (en) 2000-07-17 2004-03-30 Advanced Micro Devices, Inc. In-band management of a stacked group of switches by a single CPU
US6693906B1 (en) 2000-07-17 2004-02-17 Advanced Micro Devices, Inc. Apparatus and method for buffer-free evaluation of packet data bytes with multiple min terms
US6963565B1 (en) 2000-08-14 2005-11-08 Advanced Micro Devices, Inc. Apparatus and method for identifying data packet at wire rate on a network switch port
US6885666B1 (en) 2000-08-14 2005-04-26 Advanced Micro Devices, Inc. Apparatus and method in a network switch for synchronizing transfer of a control tag to a switch fabric with transfer of frame data to a buffer memory
US8619793B2 (en) * 2000-08-21 2013-12-31 Rockstar Consortium Us Lp Dynamic assignment of traffic classes to a priority queue in a packet forwarding device
US6407960B1 (en) 2000-09-01 2002-06-18 Advanced Micro Devices Arrangement for programming selected device registers during initialization from an external memory
US6973043B1 (en) 2000-10-20 2005-12-06 Advanced Micro Devices, Inc. Arrangement for converting between a media independent interface and a twisted pair medium using a field programmable gate array
US6564161B1 (en) 2000-12-01 2003-05-13 Advanced Micro Devices, Inc. Arrangement for testing light emitting diode interface of an integrated network device
US6634016B1 (en) 2000-12-01 2003-10-14 Advanced Micro Devices, Inc. Arrangement for partitioning logic into multiple field programmable gate arrays
US6795858B1 (en) 2000-12-29 2004-09-21 Cisco Technology, Inc. Method and apparatus for metric based server selection
US6912202B1 (en) 2001-01-25 2005-06-28 Advanced Micro Device, Inc. Arrangement for testing network switch expansion port using external logic to emulate connected expansion port
US6914892B1 (en) 2001-01-29 2005-07-05 Advanced Micro Devices, Inc. Arrangement for testing network switch expansion port data by converting to media independent interface format
US6678845B1 (en) 2001-01-29 2004-01-13 Advanced Micro Devices, Inc. Arrangement for testing programmed port registers of integrated network device by reading-back values from the port registers
US6816465B1 (en) 2001-01-29 2004-11-09 Advanced Micro Devices, Inc. Arrangement for testing pause frame response in a network switch
US6785849B1 (en) 2001-02-09 2004-08-31 Advanced Micro Devices, Inc. Arrangement for reducing a media independent interface speed in a network switch emulation system
US7024603B1 (en) * 2001-03-05 2006-04-04 Advanced Micro Devices, Inc. Arrangement for verifying that memory external to a network switch and the memory interface are free of defects
US6963536B1 (en) 2001-03-23 2005-11-08 Advanced Micro Devices, Inc. Admission control in a network device
US6954427B1 (en) * 2001-03-28 2005-10-11 Advanced Micro Devices, Inc. Method and apparatus for performing priority-based admission control
US6760277B1 (en) 2001-04-06 2004-07-06 Advanced Micro Devices, Inc. Arrangement for generating multiple clocks in field programmable gate arrays of a network test system
US7062562B1 (en) 2001-04-11 2006-06-13 Cisco Technology, Inc. Methods and apparatus for content server selection
US7080138B1 (en) 2001-04-11 2006-07-18 Cisco Technology, Inc. Methods and apparatus for content server selection
US6940865B2 (en) * 2001-04-17 2005-09-06 Atheros Communications, Inc. System and method for interleaving frames with different priorities
US6892174B1 (en) 2001-05-03 2005-05-10 Advanced Micro Devices, Inc. Arrangement for testing a network device by interfacing a low speed emulation system with high speed CPU
US7536715B2 (en) * 2001-05-25 2009-05-19 Secure Computing Corporation Distributed firewall system and method
US7881208B1 (en) 2001-06-18 2011-02-01 Cisco Technology, Inc. Gateway load balancing protocol
US7085287B1 (en) 2001-06-27 2006-08-01 Cisco Technology, Inc. Map routing technique implemented in access networks
US7349430B1 (en) 2001-06-27 2008-03-25 Cisco Technology, Inc. Addressing scheme implemented in access networks
US7227863B1 (en) 2001-11-09 2007-06-05 Cisco Technology, Inc. Methods and apparatus for implementing home agent redundancy
US7227838B1 (en) 2001-12-14 2007-06-05 Cisco Technology, Inc. Enhanced internal router redundancy
US8635305B1 (en) 2001-12-19 2014-01-21 Cisco Technology, Inc. Mechanisms for providing differentiated services within a web cache
US7720997B1 (en) 2001-12-19 2010-05-18 Cisco Technology, Inc. Path selection system
FR2837586B1 (en) * 2002-03-22 2005-03-18 St Microelectronics Sa METHOD FOR ASSOCIATING A FIRST ADDRESS WITH A SECOND ADDRESS OF REDUCED SIZE
US7251704B2 (en) * 2002-08-23 2007-07-31 Intel Corporation Store and forward switch device, system and method
US7594262B2 (en) * 2002-09-04 2009-09-22 Secure Computing Corporation System and method for secure group communications
US7308706B2 (en) * 2002-10-28 2007-12-11 Secure Computing Corporation Associative policy model
US7643502B2 (en) * 2002-12-19 2010-01-05 Intel Corporation Method and apparatus to perform frame coalescing
US7814232B2 (en) 2003-03-28 2010-10-12 Cisco Technology, Inc. Network address translation with gateway load distribution
JP4257509B2 (en) * 2003-06-27 2009-04-22 日本電気株式会社 Network system, node device, redundancy construction method, and redundancy construction program
US20040264479A1 (en) * 2003-06-30 2004-12-30 Makaram Raghunandan Method for generating a trie having a reduced number of trie blocks
US7593346B2 (en) * 2003-07-31 2009-09-22 Cisco Technology, Inc. Distributing and balancing traffic flow in a virtual gateway
US7486688B2 (en) * 2004-03-29 2009-02-03 Conexant Systems, Inc. Compact packet switching node storage architecture employing Double Data Rate Synchronous Dynamic RAM
US7840988B1 (en) 2004-05-07 2010-11-23 Cisco Technology, Inc. Front-end structure for access network line card
US8180883B1 (en) 2004-08-02 2012-05-15 Cisco Technology, Inc. Method and system for processing directives included in management events
US8059661B2 (en) * 2004-12-29 2011-11-15 Cisco Technology, Inc. Methods and apparatus for using DHCP for home address management of nodes attached to an edge device and for performing mobility and address management as a proxy home agent
US7623455B2 (en) * 2005-04-02 2009-11-24 Cisco Technology, Inc. Method and apparatus for dynamic load balancing over a network link bundle
CN101167330B (en) * 2005-04-25 2011-07-13 三星电子株式会社 Method for transmitting/receiving data in a communication system
US20070198730A1 (en) * 2006-02-17 2007-08-23 Via Technologies Inc. Embedded system with a bus and arbitration method for same
JP4834493B2 (en) * 2006-08-25 2011-12-14 アラクサラネットワークス株式会社 Network relay device and method for controlling network relay device
US8667175B2 (en) * 2008-03-13 2014-03-04 Cisco Technology, Inc. Server selection for routing content to a client using application layer redirection
CN101656586B (en) * 2008-08-20 2013-08-07 中兴通讯股份有限公司 Method and device for improving virtual concatenation delay compensation caching efficiency in synchronous digital hierarchy
US20170093730A1 (en) 2015-09-25 2017-03-30 FSA Technologies,Inc. Flow control system and method
CN113867287B (en) * 2021-09-28 2023-12-05 浙江华章科技有限公司 Industrial data acquisition method and system

Family Cites Families (11)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5231633A (en) * 1990-07-11 1993-07-27 Codex Corporation Method for prioritizing, selectively discarding, and multiplexing differing traffic type fast packets
US5193110A (en) * 1990-10-09 1993-03-09 Boston Technology, Incorporated Integrated services platform for telephone communication system
JPH05167585A (en) * 1991-12-16 1993-07-02 Nec Corp Local area network multi-address frame communication device and method
US5390299A (en) * 1991-12-27 1995-02-14 Digital Equipment Corporation System for using three different methods to report buffer memory occupancy information regarding fullness-related and/or packet discard-related information
US5276681A (en) * 1992-06-25 1994-01-04 Starlight Networks Process for fair and prioritized access to limited output buffers in a multi-port switch
JP3104429B2 (en) * 1992-10-08 2000-10-30 株式会社日立製作所 Common buffer type ATM switch having copy function and copy method thereof
GB9309468D0 (en) * 1993-05-07 1993-06-23 Roke Manor Research Improvements in or relating to asynchronous transfer mode communication systems
DE69329709D1 (en) * 1993-04-29 2001-01-04 Ibm Method and device for multiple transmission of data in a communication system
US5515376A (en) * 1993-07-19 1996-05-07 Alantec, Inc. Communication apparatus and methods
US5812792A (en) * 1994-07-22 1998-09-22 Network Peripherals, Inc. Use of video DRAM for memory storage in a local area network port of a switching hub
EP0769863B1 (en) * 1995-10-20 2004-01-14 International Business Machines Corporation Bridging apparatus for traffic filtering in communication networks

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US7899924B2 (en) 2002-04-19 2011-03-01 Oesterreicher Richard T Flexible streaming hardware

Also Published As

Publication number Publication date
JP2001511978A (en) 2001-08-14
EP0960503B1 (en) 2002-01-02
DE69803364T2 (en) 2002-10-17
JP4603102B2 (en) 2010-12-22
DE69803364D1 (en) 2002-02-28
WO1998036529A1 (en) 1998-08-20
US5953335A (en) 1999-09-14
EP0960503A1 (en) 1999-12-01

Similar Documents

Publication Publication Date Title
TW453080B (en) Method and apparatus for selectively discarding packet for blocked output queues in the network switch
TW449694B (en) Method and apparatus for transmitting multiple copies by replicating data identifiers
JP2002514366A (en) Multi-copy queue structure with searchable cache area
EP0960536B1 (en) Queuing structure and method for prioritization of frames in a network switch
EP0960511B1 (en) Method and apparatus for reclaiming buffers
JP4166285B2 (en) Method and apparatus for reclaiming a buffer
US6175902B1 (en) Method and apparatus for maintaining a time order by physical ordering in a memory
TW513635B (en) Method and structure for variable-length frame support in a shared memory switch
EP0960505B1 (en) Method and apparatus for controlling initiation of transmission of data as a function of received data
JP3652245B2 (en) Packet switch
EP0960510B1 (en) Split-queue architecture and method of queuing
TW381223B (en) Split-queue architecture and method of queuing
JP4452781B2 (en) Method and apparatus for maintaining time order by physical order in memory